Cluster role changes apply to delivery and tasks without a restart

In cluster rehearsal 3, turning outboundMta off on node1's role was
reported applied (x:settingsReload applied: true), yet node1 kept
delivering mail, a report message included, until it was restarted.
The queue and report managers were started at boot only when the
node's role included outboundMta (crates/smtp/src/lib.rs), and the task
manager only when the role had some task type (spawn_task_manager).
After that nothing looked at the role again: a queue manager that was
running kept claiming and delivering, and one that wasn't never
started.

They now start on every node (outside recovery mode) and follow the
role live:

- Queue manager: before each scan it reads the role from the running
  settings. Without outboundMta it claims nothing new; deliveries
  already running finish and report back as usual, which releases
  their locks. When the role comes back (a reload wakes the manager
  with ReloadSettings, and it looks again every 30 s regardless) it
  logs queue.started and scans the whole queue at once.
- Report scheduler: DMARC and TLS report events are handled only while
  the role has outboundMta, as at boot; events arriving without it are
  dropped, as they were on a node started without the role.
- Task manager: task_enabled already read the current role on every
  scan. It now also runs on nodes whose role has no task type (the
  scan returns at once until one is added), a job claimed before a
  role change is handed back at once rather than run or held until
  its lease lapses, and a settings reload wakes the manager so a role
  that gained task types starts claiming them straight away.

Starting the queue manager on every node also drains the queue channel
on nodes without outboundMta. Upstream left that channel unread, so
each message queued there parked a refresh in it, and by the code,
queueing would block once 1024 had piled up (not reproduced here).

A role object edit reaches the nodes that name that role in
INBUXA_ROLE. Moving a node to another role still means changing its
environment, and so a restart. Listener changes in a role still need a
restart too (listeners bind at boot); this change is about tasks and
delivery.

cluster::live_roles::live_role_tests (new; PostgreSQL, two nodes over
one store):
1. A node started with outboundMta delivers and runs a TLS report
   task; after its role loses outboundMta and the settings reload, a
   new message isn't attempted and a new report task stays pending;
   with the role back, both are taken up.
2. A node started with no task type at all gains outboundMta: a
   waiting message is attempted and a report task runs.
On main the test fails at step 1 ("delivery attempted without
outboundMta"); with step 1 bypassed, step 2 fails (nothing picked the
message up in 20 s).

crates/common/src/cache/reload.rs

@@ -13,7 +13,7 @@ use crate::{
         storage::Storage,
         telemetry::Telemetry,
     },
-    ipc::{QueueEvent, RegistryChange},
+    ipc::{BroadcastEvent, QueueEvent, RegistryChange},
     network::security::{BlockedIps, IpWithTtl},
 };
 use ahash::AHashMap;
@@ -155,6 +155,12 @@ impl Server {
                             .await
                             .ok();
 
+                        // inbuxa: the task manager reads the node's role on
+                        // every scan; scan now, so a role that gained task
+                        // types starts claiming them without waiting out the
+                        // refresh interval
+                        self.inner.ipc.task_tx.notify_one();
+
                         self.record_build_errors(&bootstrap.errors);
 
                         return Ok(ReloadResult {
@@ -232,3 +238,206 @@ fn error_object(error: &Error) -> Option<ObjectId> {
         Error::Internal { object_id, .. } => *object_id,
     }
 }
+
+// inbuxa: upstream applied a registry write to the running settings only on
+// an explicit x:Action ReloadSettings (Directory and Authentication aside), so
+// a new MtaDeliverySchedule, say, stayed unknown ("Queue strategy not found")
+// until someone reloaded. Writes to objects the settings are built from now
+// reload them, here and across the cluster, as ReloadSettings does.
+
+/// Coalesces the full reloads that registry writes trigger: a write waits for
+/// a reload that started after it was stored, and joins one if it can, so a
+/// burst of writes costs a reload or two rather than one each.
+#[derive(Default)]
+pub struct SettingsReloadGate {
+    requested: std::sync::atomic::AtomicU64,
+    state: tokio::sync::Mutex<SettingsReloadState>,
+}
+
+#[derive(Default)]
+struct SettingsReloadState {
+    completed: u64,
+    refused: Option<String>,
+}
+
+/// The reload a write to `object` calls for: the object to reload, or None
+/// when the running settings don't hold that object (accounts, domains and
+/// other data read as needed, stores, which take a restart, and objects with
+/// reload actions of their own, such as applications). Blocked IPs have a
+/// reload of their own; allowed IPs take the full one.
+pub fn write_reload_target(object: ObjectType) -> Option<ObjectType> {
+    match object {
+        ObjectType::Certificate => Some(ObjectType::Certificate),
+        ObjectType::MemoryLookupKey
+        | ObjectType::MemoryLookupKeyValue
+        | ObjectType::HttpLookup
+        | ObjectType::StoreLookup => Some(ObjectType::StoreLookup),
+        ObjectType::BlockedIp => Some(ObjectType::BlockedIp),
+        // Allowed IPs are part of the core's security settings
+        // (Security::parse), which only a full reload rebuilds; the blocked-IP
+        // reload doesn't touch them
+        ObjectType::AllowedIp
+        | ObjectType::AcmeProvider
+        | ObjectType::AddressBook
+        | ObjectType::AiModel
+        | ObjectType::Asn
+        | ObjectType::Authentication
+        | ObjectType::Cache
+        | ObjectType::Calendar
+        | ObjectType::CalendarAlarm
+        | ObjectType::CalendarScheduling
+        | ObjectType::ClusterRole
+        | ObjectType::DataRetention
+        | ObjectType::Directory
+        | ObjectType::DkimReportSettings
+        | ObjectType::DmarcReportSettings
+        | ObjectType::DnsResolver
+        | ObjectType::DsnReportSettings
+        | ObjectType::Email
+        | ObjectType::EventTracingLevel
+        | ObjectType::FileStorage
+        | ObjectType::Http
+        | ObjectType::HttpForm
+        | ObjectType::Imap
+        | ObjectType::Jmap
+        | ObjectType::Metrics
+        | ObjectType::MtaConnectionStrategy
+        | ObjectType::MtaDeliverySchedule
+        | ObjectType::MtaExtensions
+        | ObjectType::MtaHook
+        | ObjectType::MtaInboundSession
+        | ObjectType::MtaInboundThrottle
+        | ObjectType::MtaMilter
+        | ObjectType::MtaOutboundStrategy
+        | ObjectType::MtaOutboundThrottle
+        | ObjectType::MtaQueueQuota
+        | ObjectType::MtaRoute
+        | ObjectType::MtaStageAuth
+        | ObjectType::MtaStageConnect
+        | ObjectType::MtaStageData
+        | ObjectType::MtaStageEhlo
+        | ObjectType::MtaStageMail
+        | ObjectType::MtaStageRcpt
+        | ObjectType::MtaSts
+        | ObjectType::MtaTlsStrategy
+        | ObjectType::MtaVirtualQueue
+        | ObjectType::NetworkListener
+        | ObjectType::OidcProvider
+        | ObjectType::ReportSettings
+        | ObjectType::Search
+        | ObjectType::Security
+        | ObjectType::SenderAuth
+        | ObjectType::Sharing
+        | ObjectType::SieveSystemInterpreter
+        | ObjectType::SieveSystemScript
+        | ObjectType::SieveUserInterpreter
+        | ObjectType::SieveUserScript
+        | ObjectType::SpamClassifier
+        | ObjectType::SpamDnsblServer
+        | ObjectType::SpamDnsblSettings
+        | ObjectType::SpamFileExtension
+        | ObjectType::SpamPyzor
+        | ObjectType::SpamRule
+        | ObjectType::SpamSettings
+        | ObjectType::SpamTag
+        | ObjectType::SpfReportSettings
+        | ObjectType::SystemSettings
+        | ObjectType::TaskManager
+        | ObjectType::TlsReportSettings
+        | ObjectType::Tracer
+        | ObjectType::WebDav
+        | ObjectType::WebHook => Some(object),
+        _ => None,
+    }
+}
+
+impl Server {
+    /// Applies a stored registry write to `object` to the running settings,
+    /// and on success tells the other nodes to do the same. Returns None when
+    /// the write needs no reload, Some(Ok(())) when it was applied, and
+    /// Some(Err(reason)) when the reload was refused (the write stays stored;
+    /// ReloadSettings reports the same errors).
+    pub async fn reload_after_write(&self, object: ObjectType) -> Option<Result<(), String>> {
+        let target = write_reload_target(object)?;
+        let change = RegistryChange::Reload(target);
+
+        if matches!(
+            target,
+            ObjectType::Certificate | ObjectType::StoreLookup | ObjectType::BlockedIp
+        ) {
+            // Cheap, and limited to their own objects
+            let result = self.reload_and_broadcast(change).await;
+            return Some(result);
+        }
+
+        let gate = &self.inner.data.settings_reload;
+        let ticket = gate
+            .requested
+            .fetch_add(1, std::sync::atomic::Ordering::SeqCst)
+            + 1;
+        let mut state = gate.state.lock().await;
+        if state.completed >= ticket {
+            // A reload that started after this write was stored has run
+            return Some(state.refused.clone().map_or(Ok(()), Err));
+        }
+        let covers = gate.requested.load(std::sync::atomic::Ordering::SeqCst);
+        let result = self.reload_and_broadcast(change).await;
+        state.completed = covers;
+        state.refused = result.clone().err();
+        Some(result)
+    }
+
+    async fn reload_and_broadcast(&self, change: RegistryChange) -> Result<(), String> {
+        match Box::pin(self.reload_registry(change)).await {
+            Ok(reload) if !reload.has_errors() => {
+                reload.log();
+                self.cluster_broadcast(BroadcastEvent::RegistryChange(change))
+                    .await;
+                Ok(())
+            }
+            Ok(reload) => {
+                reload.log();
+                let reason = describe_reload_errors(&reload.errors);
+                trc::event!(
+                    Registry(trc::RegistryEvent::BuildWarning),
+                    Details = "Settings didn't reload after a registry write",
+                    Reason = reason.clone(),
+                );
+                Err(reason)
+            }
+            Err(err) => {
+                let reason = err.to_string();
+                trc::error!(err.details("Failed to reload settings after a registry write"));
+                Err(reason)
+            }
+        }
+    }
+}
+
+/// inbuxa: a refused reload's errors in a sentence: the first one, naming its
+/// object, and how many more there are.
+pub fn describe_reload_errors(errors: &[Error]) -> String {
+    let mut description = match errors.first() {
+        Some(Error::Build { object_id, message }) => format!("{object_id}: {message}"),
+        Some(Error::Validation { object_id, errors }) => format!(
+            "{object_id}: {}",
+            errors
+                .iter()
+                .map(|err| err.to_string())
+                .collect::<Vec<_>>()
+                .join("; ")
+        ),
+        Some(Error::Internal {
+            object_id: Some(object_id),
+            error,
+        }) => format!("{object_id}: {error}"),
+        Some(Error::Internal { error, .. }) => error.to_string(),
+        Some(Error::NotFound { object_id }) => format!("{object_id} was not found"),
+        None => String::new(),
+    };
+    let more = errors.len().saturating_sub(1);
+    if more > 0 {
+        description.push_str(&format!(" ({more} more in the server log.)"));
+    }
+    description
+}

crates/common/src/config/inner.rs

@@ -93,6 +93,8 @@ impl Data {
             registry_id_gen: id_generator.clone(),
             span_id_gen: id_generator,
             queue_status: true.into(),
+            settings_reload: Default::default(),
+            store_health: Default::default(),
             applications,
             logos: Default::default(),
             smtp_connectors: TlsConnectors::try_new().failed("Failed to build TLS connectors"),
@@ -235,6 +237,8 @@ impl Default for Data {
             span_id_gen: Default::default(),
             registry_id_gen: Default::default(),
             queue_status: true.into(),
+            settings_reload: Default::default(),
+            store_health: Default::default(),
             applications: WebApplications::new(),
             logos: Default::default(),
             smtp_connectors: TlsConnectors::try_new().unwrap(),

crates/common/src/ipc.rs

@@ -345,8 +345,13 @@ pub struct TaskLocks {
 }
 
 impl TaskLocks {
-    /// How long a task lock lasts, in seconds, unless it is released first.
+    /// How long a task lock lasts, in seconds, unless it is released first
-    pub const DEFAULT_EXPIRY: u64 = 60 * 60;
+    /// or renewed. inbuxa: upstream held a lock for an hour, so a killed
+    /// node's tasks waited that long; the lock is now a five-minute lease
+    /// that the task manager renews every third of it while the task runs
+    /// (renew_task_locks), so a dead node's tasks run elsewhere within
+    /// minutes.
+    pub const DEFAULT_EXPIRY: u64 = 5 * 60;
 
     pub fn is_stopping(&self) -> bool {
         self.stopping.load(Ordering::Acquire)
@@ -370,6 +375,16 @@ impl TaskLocks {
         self.held.lock().len()
     }
 
+    /// inbuxa: the tasks this node holds, to renew their locks.
+    pub fn held_ids(&self) -> Vec<u64> {
+        self.held.lock().iter().copied().collect()
+    }
+
+    /// inbuxa: whether this node holds (and is running) the task.
+    pub fn is_held(&self, id: u64) -> bool {
+        self.held.lock().contains(&id)
+    }
+
     pub fn expiry(&self) -> u64 {
         self.expiry.load(Ordering::Relaxed)
     }

crates/common/src/lib.rs

@@ -161,6 +161,10 @@ pub struct Data {
     pub span_id_gen: SnowflakeIdGenerator,
     pub registry_id_gen: SnowflakeIdGenerator,
     pub queue_status: AtomicBool,
+    // inbuxa: coalesces the settings reloads registry writes trigger
+    pub settings_reload: cache::reload::SettingsReloadGate,
+    // inbuxa: the readiness probe's cached answer
+    pub store_health: storage::ready::StoreHealth,
 
     pub applications: WebApplications,
     pub logos: Mutex<AHashMap<Box<str>, LogoCache>>,

crates/common/src/storage/mod.rs

@@ -26,6 +26,7 @@ pub mod document;
 pub mod encryption;
 pub mod index;
 pub mod quota;
+pub mod ready; // inbuxa: readiness follows the data store
 pub mod state;
 pub mod transaction;
 

crates/common/src/storage/ready.rs

@@ -0,0 +1,83 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! Readiness that reflects the data store.
+//!
+//! /healthz/ready used to answer 200 whenever a data store was configured,
+//! so a load balancer kept sending traffic to a node through a database
+//! outage. It now reads one key from the data store, with a short time
+//! limit, and caches the answer for a couple of seconds so probes can't load
+//! the database. Liveness stays 200: restarting a node doesn't bring its
+//! database back, and an orchestrator that restarts on failed liveness would
+//! otherwise restart every node at once.
+
+use crate::Server;
+use parking_lot::Mutex;
+use std::{
+    sync::atomic::{AtomicBool, Ordering},
+    time::{Duration, Instant},
+};
+use store::{ValueKey, write::ValueClass};
+
+/// How long a probe's answer is reused.
+pub const READY_CACHE: Duration = Duration::from_secs(2);
+/// How long a probe waits for the data store.
+pub const READY_PROBE_TIMEOUT: Duration = Duration::from_secs(2);
+
+#[derive(Default)]
+pub struct StoreHealth {
+    last: Mutex<Option<(Instant, bool)>>,
+    probing: AtomicBool,
+}
+
+/// Clears the probing flag even when the request is dropped mid-probe.
+struct ProbeGuard<'x>(&'x AtomicBool);
+
+impl Drop for ProbeGuard<'_> {
+    fn drop(&mut self) {
+        self.0.store(false, Ordering::Release);
+    }
+}
+
+impl Server {
+    /// Whether the data store answers: a cached result younger than
+    /// READY_CACHE, or a fresh read bounded by READY_PROBE_TIMEOUT. While
+    /// one probe is running, other callers get the last answer.
+    pub async fn is_data_store_ready(&self) -> bool {
+        let store = &self.core.storage.data;
+        if store.is_none() {
+            return false;
+        }
+        let health = &self.inner.data.store_health;
+        let last = *health.last.lock();
+        if let Some((at, ready)) = last
+            && at.elapsed() < READY_CACHE
+        {
+            return ready;
+        }
+        if health.probing.swap(true, Ordering::AcqRel) {
+            return last.is_none_or(|(_, ready)| ready);
+        }
+        let _guard = ProbeGuard(&health.probing);
+
+        let ready = tokio::time::timeout(
+            READY_PROBE_TIMEOUT,
+            store.get_value::<u64>(ValueKey::from(ValueClass::Property(0))),
+        )
+        .await
+        .is_ok_and(|result| result.is_ok());
+        // Say so once per outage, not on every probe
+        if !ready && last.is_none_or(|(_, ready)| ready) {
+            trc::event!(
+                Store(trc::StoreEvent::UnexpectedError),
+                Details = "Readiness probe: the data store didn't answer",
+                Limit = READY_PROBE_TIMEOUT,
+            );
+        }
+        *health.last.lock() = Some((Instant::now(), ready));
+        ready
+    }
+}

crates/http/src/request.rs

@@ -553,8 +553,10 @@ impl ParseHttp for Server {
                         return Ok(JsonProblemResponse(StatusCode::OK).into_http_response());
                     }
                     "ready" => {
+                        // inbuxa: ready only while the data store answers
+                        // (a cached, time-limited read); liveness stays 200
                         return Ok(JsonProblemResponse({
-                            if !self.core.storage.data.is_none() {
+                            if self.is_data_store_ready().await {
                                 StatusCode::OK
                             } else {
                                 StatusCode::SERVICE_UNAVAILABLE

crates/jmap-proto/src/method/set.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use super::ahash_is_empty;
@@ -71,6 +73,23 @@ pub struct SetResponse<T: JmapObject> {
     #[serde(rename = "notDestroyed")]
     #[serde(skip_serializing_if = "VecMap::is_empty")]
     pub not_destroyed: VecMap<MaybeInvalid<Id>, SetError<T::Property>>,
+
+    // inbuxa: on a registry write that changes the running settings, whether
+    // the server applied it
+    #[serde(rename = "x:settingsReload")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub settings_reload: Option<SettingsReload>,
+}
+
+/// inbuxa: the settings reload that followed a registry write.
+#[derive(Debug, Clone, serde::Serialize)]
+pub struct SettingsReload {
+    /// The running settings (here and, through the cluster, on every node)
+    /// include the write.
+    pub applied: bool,
+    /// Why they don't, when they don't.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub description: Option<String>,
 }
 
 impl<'de, T: JmapObject> DeserializeArguments<'de> for SetRequest<'de, T> {
@@ -199,6 +218,7 @@ impl<T: JmapObject> SetResponse<T> {
                 not_created: VecMap::new(),
                 not_updated: VecMap::new(),
                 not_destroyed: VecMap::new(),
+                settings_reload: None,
             })
         } else {
             Err(trc::JmapEvent::RequestTooLarge.into_err())

crates/jmap/src/registry/mapping/action.rs

@@ -580,29 +580,9 @@ async fn dmarc_troubleshoot(
 /// settings weren't applied; upstream passed on the first error's bare message
 /// ("Invalid address: ..."), which read like a problem with the request.
 fn reload_refused(errors: Vec<registry::types::error::Error>) -> SetError<Property> {
-    use registry::types::error::Error;
+    let description = format!(
-    let more = errors.len().saturating_sub(1);
+        "Settings were not reloaded. {}",
-    let mut description = match errors.first() {
+        common::cache::reload::describe_reload_errors(&errors)
-        Some(Error::Build { object_id, message }) => format!("{object_id}: {message}"),
+    );
-        Some(Error::Validation { object_id, errors }) => format!(
-            "{object_id}: {}",
-            errors
-                .iter()
-                .map(|err| err.to_string())
-                .collect::<Vec<_>>()
-                .join("; ")
-        ),
-        Some(Error::Internal {
-            object_id: Some(object_id),
-            error,
-        }) => format!("{object_id}: {error}"),
-        Some(Error::Internal { error, .. }) => error.to_string(),
-        Some(Error::NotFound { object_id }) => format!("{object_id} was not found"),
-        None => String::new(),
-    };
-    description.insert_str(0, "Settings were not reloaded. ");
-    if more > 0 {
-        description.push_str(&format!(" ({more} more in the server log.)"));
-    }
     map_bootstrap_error(errors).with_description(description)
 }

crates/jmap/src/registry/set.rs

@@ -38,7 +38,7 @@ use directory::core::secret::{hash_secret, is_password_hash};
 use http_proto::HttpSessionData;
 use jmap_proto::{
     error::set::{SetError, SetErrorType},
-    method::set::{SetRequest, SetResponse},
+    method::set::{SetRequest, SetResponse, SettingsReload},
     object::registry::Registry,
     references::resolve::ResolveCreatedReference,
     request::{IntoValid, MaybeInvalid},
@@ -931,34 +931,28 @@ impl RegistrySet for Server {
             }
         };
 
-        // inbuxa: DIR-17: a directory or the server default applies on the
+        // inbuxa: a write to an object the running settings are built from
-        // next request, here and on every node
+        // applies at once, here and on every node (DIR-17 did this for
-        if matches!(
+        // directories and the server default; now it covers every such object)
-            object_type,
+        let mut result = result;
-            ObjectType::Directory | ObjectType::Authentication
+        if let Ok(response) = &mut result
-        ) && let Ok(response) = &result
             && (!response.created.is_empty()
                 || !response.updated.is_empty()
                 || !response.destroyed.is_empty())
+            && let Some(reload) = self.reload_after_write(object_type).await
         {
-            let change = common::ipc::RegistryChange::Reload(ObjectType::Directory);
+            response.settings_reload = Some(match reload {
-            match Box::pin(self.reload_registry(change)).await {
+                Ok(()) => SettingsReload {
-                Ok(reload) if !reload.has_errors() => {
+                    applied: true,
-                    self.cluster_broadcast(common::ipc::BroadcastEvent::RegistryChange(change))
+                    description: None,
-                        .await;
+                },
-                }
+                Err(reason) => SettingsReload {
-                Ok(reload) => {
+                    applied: false,
-                    // inbuxa: name what stopped it
+                    description: Some(format!(
-                    reload.log();
+                        "Saved, but the running settings were not reloaded. {reason}"
-                    trc::event!(
+                    )),
-                        Registry(trc::RegistryEvent::BuildWarning),
+                },
-                        Details = "Settings didn't reload after a directory change",
+            });
-                    )
-                }
-                Err(err) => {
-                    trc::error!(err.details("Failed to reload directories"));
-                }
-            }
         }
         result
     }

crates/services/src/task_manager/lock.rs

@@ -82,3 +82,42 @@ pub async fn release_task_locks(server: &Server) -> usize {
     }
     ids.len()
 }
+
+/// inbuxa: renews the lease on every task this node is running, so it stays
+/// claimed for as long as it runs while a node that dies loses its claims
+/// within one lock lifetime. Returns how many leases were renewed and how
+/// many were found lost (expired, perhaps taken by another node).
+pub async fn renew_task_locks(server: &Server) -> (usize, usize) {
+    let locks = &server.inner.ipc.task_locks;
+    let expiry = locks.expiry();
+    let (mut renewed, mut lost) = (0, 0);
+    for id in locks.held_ids() {
+        match server
+            .in_memory_store()
+            .renew_lock(KV_LOCK_TASK, &id.to_be_bytes(), expiry)
+            .await
+        {
+            Ok(true) => renewed += 1,
+            Ok(false) => {
+                // Still held here as far as this node knows; the task
+                // finishes and its lock is removed as usual
+                if locks.is_held(id) {
+                    lost += 1;
+                    trc::event!(
+                        TaskManager(TaskManagerEvent::TaskLocked),
+                        Id = id,
+                        Details = "Task lock expired while the task was running",
+                    );
+                }
+            }
+            Err(err) => {
+                trc::error!(
+                    err.details("Failed to renew task lock")
+                        .ctx(trc::Key::Id, id)
+                        .caused_by(trc::location!())
+                );
+            }
+        }
+    }
+    (renewed, lost)
+}

crates/services/src/task_manager/manager.rs

@@ -13,7 +13,7 @@ use crate::task_manager::dkim::DkimManagementTask;
 use crate::task_manager::dns::DnsManagementTask;
 use crate::task_manager::imip::SendImipTask;
 use crate::task_manager::index::SearchIndexTask;
-use crate::task_manager::lock::TaskLockManager;
+use crate::task_manager::lock::{TaskLockManager, renew_task_locks};
 use crate::task_manager::maintenance::MaintenanceTask;
 use crate::task_manager::merge_threads::MergeThreadsTask;
 use crate::task_manager::report::{self, SubmitReportTask};
@@ -24,6 +24,7 @@ use crate::task_manager::{
     TaskJob, TaskManagerIpc, TaskResult,
 };
 use common::BuildServer;
+use common::config::network::ClusterRoles;
 use common::config::server::{DEFAULT_TLS_TIMEOUT, ServerProtocol};
 use common::network::limiter::ConcurrencyLimiter;
 use common::network::{ServerInstance, TcpAcceptor};
@@ -54,24 +55,37 @@ const PERPETUAL_RETRY_MIN_DELAY: u64 = 3600;
 const PERPETUAL_RETRY_MAX_DELAY: u64 = 21600;
 
 pub fn spawn_task_manager(inner: Arc<Inner>) {
-    let is_clustered = {
+    // inbuxa: upstream didn't start the task manager on a node whose role
-        let server = inner.build_server();
+    // had no task types at boot, so adding one later did nothing until a
-        let roles = &server.core.network.roles;
+    // restart. It now always runs and reads the role on every scan and
-
+    // before every job (task_enabled), so a role change applies at the next
-        if !roles.account_maintenance
+    // settings reload.
-            && !roles.store_maintenance
+    let is_clustered = inner.build_server().core.storage.coordinator.is_enabled();
-            && !roles.search_indexing
-            && !roles.spam_training
-            && !roles.task_manager
-        {
-            return;
-        }
-
-        server.core.storage.coordinator.is_enabled()
-    };
 
     trc::event!(TaskManager(TaskManagerEvent::ManagerStarted));
 
+    // inbuxa: keep the leases of running tasks alive, every third of a lock
+    // lifetime, until the node stops
+    {
+        let inner = inner.clone();
+        tokio::spawn(async move {
+            let mut renewed_at = Instant::now();
+            loop {
+                tokio::time::sleep(Duration::from_secs(1)).await;
+                let locks = &inner.ipc.task_locks;
+                if locks.is_stopping() {
+                    break;
+                }
+                if renewed_at.elapsed() >= Duration::from_secs((locks.expiry() / 3).max(1)) {
+                    renewed_at = Instant::now();
+                    if locks.held() > 0 {
+                        renew_task_locks(&inner.build_server()).await;
+                    }
+                }
+            }
+        });
+    }
+
     // Create dummy server instance for alarms
     let server_instance = Arc::new(ServerInstance {
         id: "_local".to_string(),
@@ -126,20 +140,23 @@ pub fn spawn_task_manager(inner: Arc<Inner>) {
                     let server = inner.build_server();
                     let batch_size = server.core.email.index_batch_size;
                     let mut batch = Vec::with_capacity(batch_size);
-                    if let Some(task) = fetch_task(&server, job).await {
+                    if let Some(task) = fetch_enabled_task(&server, job).await {
                         batch.push(task);
                     }
 
                     while batch.len() < batch_size {
                         match rx.try_recv() {
                             Ok(job) => {
-                                if let Some(task) = fetch_task(&server, job).await {
+                                if let Some(task) = fetch_enabled_task(&server, job).await {
                                     batch.push(task);
                                 }
                             }
                             Err(_) => break,
                         }
                     }
+                    if batch.is_empty() {
+                        continue;
+                    }
 
                     // Dispatch. inbuxa: on a task of its own, so a panic
                     // releases the batch's locks and leaves this worker
@@ -180,7 +197,8 @@ pub fn spawn_task_manager(inner: Arc<Inner>) {
                     let server = inner.build_server();
                     let mut refresh_queue = false;
 
-                    if let Some(TaskDetails { task, info }) = fetch_task(&server, job).await {
+                    if let Some(TaskDetails { task, info }) = fetch_enabled_task(&server, job).await
+                    {
                         // inbuxa: on a task of its own, as above
                         let run = {
                             let server = server.clone();
@@ -249,6 +267,17 @@ impl TaskQueueManager for Server {
         if task_locks.is_stopping() {
             return Duration::from_secs(QUEUE_REFRESH_INTERVAL);
         }
+        // inbuxa: with no task type enabled by this node's role there is
+        // nothing to claim; a settings reload wakes the manager when that
+        // changes
+        let roles = &self.core.network.roles;
+        if !(0..TaskType::COUNT as u16)
+            .filter_map(TaskType::from_id)
+            .any(|task_type| task_enabled(roles, task_type))
+        {
+            ipc.locked.clear();
+            return Duration::from_secs(QUEUE_REFRESH_INTERVAL);
+        }
         let lock_expiry = task_locks.expiry();
         let now_timestamp = now();
         let from_key = ValueKey::<ValueClass> {
@@ -271,7 +300,6 @@ impl TaskQueueManager for Server {
         let mut tasks = Vec::new();
         let now = Instant::now();
         let mut next_event = None;
-        let roles = &self.core.network.roles;
         ipc.revision += 1;
         let _ = self
             .store()
@@ -289,26 +317,13 @@ impl TaskQueueManager for Server {
                                     .caused_by(trc::location!())
                                     .ctx(trc::Key::Value, value)
                             })?;
-                            let enabled = match task_type {
+                            // inbuxa: running here under a lease this node
-                                TaskType::IndexDocument
+                            // renews; don't hand it to a worker again
-                                | TaskType::UnindexDocument
+                            if task_locks.is_held(task_id) {
-                                | TaskType::IndexTrace => roles.search_indexing,
+                                return Ok(true);
-                                TaskType::AccountMaintenance
+                            }
-                                | TaskType::TenantMaintenance
+
-                                | TaskType::DestroyAccount => roles.account_maintenance,
+                            let enabled = task_enabled(roles, task_type);
-                                TaskType::StoreMaintenance => roles.store_maintenance,
-                                TaskType::SpamFilterMaintenance => roles.spam_training,
-                                TaskType::CalendarAlarmEmail
-                                | TaskType::CalendarAlarmNotification
-                                | TaskType::CalendarItipMessage
-                                | TaskType::MergeThreads
-                                | TaskType::DmarcReport
-                                | TaskType::TlsReport
-                                | TaskType::RestoreArchivedItem
-                                | TaskType::AcmeRenewal
-                                | TaskType::DkimManagement
-                                | TaskType::DnsManagement => true,
-                            };
 
                             if !enabled {
                                 trc::event!(
@@ -437,6 +452,48 @@ impl TaskQueueManager for Server {
     }
 }
 
+/// inbuxa: whether this node's cluster role lets it run a task type. Upstream
+/// checked the dedicated roles (search indexing, account and store
+/// maintenance, spam training) and let every node with a task manager run
+/// the rest, whatever its taskQueueProcessing setting. Every task type now
+/// answers to one ClusterTaskType:
+///
+/// - IndexDocument, UnindexDocument, IndexTrace: searchIndexing
+/// - AccountMaintenance, TenantMaintenance, DestroyAccount: accountMaintenance
+/// - StoreMaintenance: storeMaintenance
+/// - SpamFilterMaintenance: spamClassifierTraining
+/// - DmarcReport, TlsReport: outboundMta. They build and send reports to
+///   other domains (TLS reports can go straight to an HTTPS endpoint), which
+///   is the outbound MTA's business.
+/// - CalendarAlarmEmail, CalendarAlarmNotification, CalendarItipMessage,
+///   MergeThreads, RestoreArchivedItem, AcmeRenewal, DkimManagement,
+///   DnsManagement: taskQueueProcessing, the role for queue tasks with no
+///   role of their own.
+///
+/// A node that may not run a task leaves it unclaimed, so a node that may
+/// picks it up.
+pub fn task_enabled(roles: &ClusterRoles, task_type: TaskType) -> bool {
+    match task_type {
+        TaskType::IndexDocument | TaskType::UnindexDocument | TaskType::IndexTrace => {
+            roles.search_indexing
+        }
+        TaskType::AccountMaintenance | TaskType::TenantMaintenance | TaskType::DestroyAccount => {
+            roles.account_maintenance
+        }
+        TaskType::StoreMaintenance => roles.store_maintenance,
+        TaskType::SpamFilterMaintenance => roles.spam_training,
+        TaskType::DmarcReport | TaskType::TlsReport => roles.outbound_mta,
+        TaskType::CalendarAlarmEmail
+        | TaskType::CalendarAlarmNotification
+        | TaskType::CalendarItipMessage
+        | TaskType::MergeThreads
+        | TaskType::RestoreArchivedItem
+        | TaskType::AcmeRenewal
+        | TaskType::DkimManagement
+        | TaskType::DnsManagement => roles.task_manager,
+    }
+}
+
 async fn run_task(
     server: &Server,
     task: &Task,
@@ -490,6 +547,25 @@ async fn run_task(
     }
 }
 
+/// inbuxa: reads a claimed task when this node's role still allows its type.
+/// The role may have changed since the task was claimed (a settings reload in
+/// between); the claim is then handed back at once for a node that may run
+/// it, rather than held until the lease runs out.
+async fn fetch_enabled_task(server: &Server, job: TaskJob) -> Option<TaskDetails> {
+    if task_enabled(&server.core.network.roles, job.typ) {
+        fetch_task(server, job).await
+    } else {
+        trc::event!(
+            TaskManager(TaskManagerEvent::TaskIgnored),
+            Id = job.id,
+            Details = job.typ.as_str(),
+            Reason = "Task type was disabled by cluster roles after it was claimed.",
+        );
+        server.remove_index_lock(job.id).await;
+        None
+    }
+}
+
 /// Reads a claimed task. When it is gone or can't be read, the claim is
 /// released: inbuxa: holding it would block the task, everywhere, until
 /// the lock expired.

crates/smtp/src/lib.rs

@@ -44,7 +44,14 @@ impl StartQueueManager for BootManager {
 impl SpawnQueueManager for IpcReceivers {
     fn spawn_queue_manager(&mut self, inner: Arc<Inner>) {
         let core = inner.shared_core.load();
-        if !core.storage.registry.is_recovery_mode() && core.network.roles.outbound_mta {
+        // inbuxa: upstream started these only when the node's role included
+        // outboundMta at boot, so turning the role on later did nothing and
+        // turning it off left them delivering until a restart. They now run
+        // on every node and follow the role live (see Queue::start and the
+        // report scheduler). This also drains the queue channel on nodes
+        // without the role, where every queued message's refresh used to sit
+        // in a channel nobody read until it filled and queueing blocked.
+        if !core.storage.registry.is_recovery_mode() {
             // Spawn queue manager
             self.queue_rx.take().unwrap().spawn(inner.clone());
 

crates/smtp/src/queue/manager.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use super::{Message, QueueId, Status, spool::SmtpSpool};
@@ -39,6 +41,9 @@ pub struct Queue {
     pub urgent_refresh: bool,
     pub last_scan: Instant,
     pub last_full_scan: Instant,
+    /// inbuxa: whether this node's role included outboundMta when last
+    /// checked (None before the first check)
+    pub role_enabled: Option<bool>,
 }
 
 #[derive(Debug)]
@@ -67,6 +72,9 @@ impl SpawnQueue for mpsc::Receiver<QueueEvent> {
 const BACK_PRESSURE_WARN_INTERVAL: Duration = Duration::from_secs(60);
 const MIN_SCAN_INTERVAL: Duration = Duration::from_millis(100);
 const FULL_SCAN_INTERVAL: Duration = Duration::from_secs(QUEUE_REFRESH / 2);
+/// inbuxa: how often a node without the outbound MTA role looks at its role
+/// again when nothing else wakes it (a settings reload does)
+const ROLE_RECHECK_INTERVAL: Duration = Duration::from_secs(30);
 
 impl Queue {
     pub fn new(core: Arc<Inner>, rx: mpsc::Receiver<QueueEvent>) -> Self {
@@ -87,6 +95,7 @@ impl Queue {
             urgent_refresh: false,
             last_scan: now.checked_sub(MIN_SCAN_INTERVAL).unwrap_or(now),
             last_full_scan: now,
+            role_enabled: None,
         }
     }
 
@@ -123,6 +132,27 @@ impl Queue {
                 continue;
             }
 
+            // inbuxa: follow the node's role live. Without outboundMta the
+            // queue claims nothing new; deliveries already running finish
+            // and report back as usual, releasing their locks. When the role
+            // comes back, the whole queue is scanned at once.
+            let role_enabled = self.core.shared_core.load().network.roles.outbound_mta;
+            if self.role_enabled.replace(role_enabled) == Some(false) && role_enabled {
+                trc::event!(
+                    Queue(trc::QueueEvent::Started),
+                    Details = "This node's cluster role now includes outboundMta",
+                );
+                self.scan_from = 0;
+                self.pending_refresh = true;
+                self.urgent_refresh = true;
+            }
+            if !role_enabled {
+                self.pending_refresh = false;
+                self.urgent_refresh = false;
+                self.next_refresh = Instant::now() + ROLE_RECHECK_INTERVAL;
+                continue;
+            }
+
             self.pending_refresh |= refresh_queue;
             if !self.pending_refresh && self.next_refresh > Instant::now() {
                 continue;

crates/smtp/src/reporting/scheduler.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use super::{dmarc::DmarcReporting, tls::TlsReporting};
@@ -18,6 +20,14 @@ impl SpawnReport for mpsc::Receiver<ReportingEvent> {
         tokio::spawn(async move {
             while let Some(event) = self.recv().await {
                 let server = inner.build_server();
+                // inbuxa: reports are the outbound MTA's business, as at
+                // boot, but the role is read per event so a change applies
+                // without a restart. Events that arrive while the role is
+                // off are dropped, as they were on a node started without it
+                if !matches!(event, ReportingEvent::Stop) && !server.core.network.roles.outbound_mta
+                {
+                    continue;
+                }
                 match event {
                     ReportingEvent::Dmarc(event) => server.schedule_dmarc(event).await,
                     ReportingEvent::Tls(event) => server.schedule_tls(event).await,

crates/store/src/backend/mod.rs

@@ -30,6 +30,9 @@ pub mod s3;
 pub mod sqlite;
 // inbuxa: scale-out storage (sharded stores)
 pub mod scaleout;
+// inbuxa: client-side SQL query limits
+#[cfg(any(feature = "postgres", feature = "mysql"))]
+pub mod query_timeout;
 
 
 pub const MAX_TOKEN_LENGTH: usize = (u8::MAX >> 1) as usize;

crates/store/src/backend/mysql/blob.rs

@@ -2,13 +2,15 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use std::ops::Range;
 
 use mysql_async::prelude::Queryable;
 
-use super::{MysqlStore, into_error};
+use super::{MysqlStore, bounded, into_error};
 
 impl MysqlStore {
     pub(crate) async fn get_blob(
@@ -16,49 +18,64 @@ impl MysqlStore {
         key: &[u8],
         range: Range<usize>,
     ) -> trc::Result<Option<Vec<u8>>> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prep("SELECT v FROM t WHERE k = ?")
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prep("SELECT v FROM t WHERE k = ?")
-        conn.exec_first::<Vec<u8>, _, _>(&s, (key,))
+                .await
-            .await
+                .map_err(into_error)?;
-            .map(|bytes| {
+            conn.exec_first::<Vec<u8>, _, _>(&s, (key,))
-                if range.start == 0 && range.end == usize::MAX {
+                .await
-                    bytes
+                .map(|bytes| {
-                } else {
+                    if range.start == 0 && range.end == usize::MAX {
-                    bytes.map(|bytes| {
                         bytes
-                            .get(range.start..std::cmp::min(bytes.len(), range.end))
+                    } else {
-                            .unwrap_or_default()
+                        bytes.map(|bytes| {
-                            .to_vec()
+                            bytes
-                    })
+                                .get(range.start..std::cmp::min(bytes.len(), range.end))
-                }
+                                .unwrap_or_default()
-            })
+                                .to_vec()
-            .map_err(into_error)
+                        })
+                    }
+                })
+                .map_err(into_error)
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn put_blob(&self, key: &[u8], data: &[u8]) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prep("INSERT INTO t (k, v) VALUES (?, ?) ON DUPLICATE KEY UPDATE v = VALUES(v)")
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prep("INSERT INTO t (k, v) VALUES (?, ?) ON DUPLICATE KEY UPDATE v = VALUES(v)")
-        conn.exec_drop(&s, (key, data))
+                .await
-            .await
+                .map_err(into_error)?;
-            .map_err(into_error)
+            conn.exec_drop(&s, (key, data))
-            .map(|_| ())
+                .await
+                .map_err(into_error)
+                .map(|_| ())
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn delete_blob(&self, key: &[u8]) -> trc::Result<bool> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prep("DELETE FROM t WHERE k = ?")
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prep("DELETE FROM t WHERE k = ?")
-        conn.exec_iter(&s, (key,))
+                .await
-            .await
+                .map_err(into_error)?;
-            .map_err(into_error)
+            conn.exec_iter(&s, (key,))
-            .map(|hits| hits.affected_rows() > 0)
+                .await
+                .map_err(into_error)
+                .map(|hits| hits.affected_rows() > 0)
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }

crates/store/src/backend/mysql/lookup.rs

@@ -2,13 +2,15 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use mysql_async::{Params, Row, prelude::Queryable};
 
 use crate::{IntoRows, QueryResult, QueryType, Value};
 
-use super::{MysqlStore, into_error};
+use super::{MysqlStore, bounded, into_error};
 
 impl MysqlStore {
     pub(crate) async fn sql_query<T: QueryResult>(
@@ -16,28 +18,33 @@ impl MysqlStore {
         query: &str,
         params: &[Value<'_>],
     ) -> trc::Result<T> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn.prep(query).await.map_err(into_error)?;
+        let limit = self.timeouts.query;
-        let params = Params::Positional(params.iter().map(Into::into).collect());
+        let result = tokio::time::timeout(limit, async {
+            let s = conn.prep(query).await.map_err(into_error)?;
+            let params = Params::Positional(params.iter().map(Into::into).collect());
 
-        match T::query_type() {
+            match T::query_type() {
-            QueryType::Execute => conn.exec_drop(s, params).await.map_or_else(
+                QueryType::Execute => conn.exec_drop(s, params).await.map_or_else(
-                |e| Err(into_error(e)),
+                    |e| Err(into_error(e)),
-                |_| Ok(T::from_exec(conn.affected_rows() as usize)),
+                    |_| Ok(T::from_exec(conn.affected_rows() as usize)),
-            ),
+                ),
-            QueryType::Exists => conn
+                QueryType::Exists => conn
-                .exec_first::<Row, _, _>(s, params)
+                    .exec_first::<Row, _, _>(s, params)
-                .await
+                    .await
-                .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_exists(r.is_some()))),
+                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_exists(r.is_some()))),
-            QueryType::QueryOne => conn
+                QueryType::QueryOne => conn
-                .exec_first::<Row, _, _>(s, params)
+                    .exec_first::<Row, _, _>(s, params)
-                .await
+                    .await
-                .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_one(r))),
+                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_one(r))),
-            QueryType::QueryAll => conn
+                QueryType::QueryAll => conn
-                .exec::<Row, _, _>(s, params)
+                    .exec::<Row, _, _>(s, params)
-                .await
+                    .await
-                .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_all(r))),
+                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_all(r))),
-        }
+            }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }
 

crates/store/src/backend/mysql/main.rs

@@ -6,7 +6,7 @@
  * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use super::{MysqlStore, into_error};
+use super::{MysqlStore, bounded, into_error};
 use crate::{
     backend::mysql::MysqlSearchField,
     search::{
@@ -32,6 +32,9 @@ impl MysqlStore {
             .max_allowed_packet(config.max_allowed_packet.map(|v| v as usize))
             .wait_timeout(config.timeout.map(|t| t.as_secs() as usize))
             .client_found_rows(true)
+            // inbuxa: notice a server that went away without closing the
+            // connection in minutes, not the system default of two hours
+            .tcp_keepalive(Some(super::POOL_KEEPALIVE_IDLE))
             .tcp_port(config.port as u16);
 
         if config.use_tls {
@@ -69,6 +72,7 @@ impl MysqlStore {
                             .db_name(Some(replica.database.clone()))
                             .tcp_port(replica.port as u16),
                     ),
+                    timeouts: Default::default(),
                 })),
                 replica.host,
                 replica.port as u16,
@@ -78,6 +82,7 @@ impl MysqlStore {
 
         let primary = Store::MySQL(Arc::new(MysqlStore {
             conn_pool: Pool::new(opts),
+            timeouts: Default::default(),
         }));
 
         // ST-1: no replicas, no change
@@ -95,89 +100,97 @@ impl MysqlStore {
     }
 
     pub(crate) async fn create_storage_tables(&self) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
+        let limit = self.timeouts.maintenance;
+        let result = tokio::time::timeout(limit, async {
+            for table in [
+                SUBSPACE_ACL,
+                SUBSPACE_TASK_QUEUE,
+                SUBSPACE_DELETED_ITEMS,
+                SUBSPACE_SPAM_SAMPLES,
+                crate::SUBSPACE_INBUXA, // inbuxa: masked email
+                SUBSPACE_BLOB_LINK,
+                SUBSPACE_IN_MEMORY_VALUE,
+                SUBSPACE_PROPERTY,
+                SUBSPACE_REGISTRY,
+                SUBSPACE_REGISTRY_PK,
+                SUBSPACE_DIRECTORY,
+                SUBSPACE_QUEUE_MESSAGE,
+                SUBSPACE_QUEUE_EVENT,
+                SUBSPACE_REPORT_OUT,
+                SUBSPACE_REPORT_IN,
+                SUBSPACE_LOGS,
+                SUBSPACE_TELEMETRY_SPAN,
+                SUBSPACE_TELEMETRY_METRIC,
+            ] {
+                let table = char::from(table);
+                conn.query_drop(format!(
+                    "CREATE TABLE IF NOT EXISTS {table} (
+                        k VARBINARY(255) NOT NULL,
+                        v MEDIUMBLOB NOT NULL,
+                        PRIMARY KEY (k)
+                    ) ENGINE=InnoDB"
+                ))
+                .await
+                .map_err(into_error)?;
+            }
 
-        for table in [
-            SUBSPACE_ACL,
-            SUBSPACE_TASK_QUEUE,
-            SUBSPACE_DELETED_ITEMS,
-            SUBSPACE_SPAM_SAMPLES,
-            crate::SUBSPACE_INBUXA, // inbuxa: masked email
-            SUBSPACE_BLOB_LINK,
-            SUBSPACE_IN_MEMORY_VALUE,
-            SUBSPACE_PROPERTY,
-            SUBSPACE_REGISTRY,
-            SUBSPACE_REGISTRY_PK,
-            SUBSPACE_DIRECTORY,
-            SUBSPACE_QUEUE_MESSAGE,
-            SUBSPACE_QUEUE_EVENT,
-            SUBSPACE_REPORT_OUT,
-            SUBSPACE_REPORT_IN,
-            SUBSPACE_LOGS,
-            SUBSPACE_TELEMETRY_SPAN,
-            SUBSPACE_TELEMETRY_METRIC,
-        ] {
-            let table = char::from(table);
-            conn.query_drop(format!(
-                "CREATE TABLE IF NOT EXISTS {table} (
-                    k VARBINARY(255) NOT NULL,
-                    v MEDIUMBLOB NOT NULL,
-                    PRIMARY KEY (k)
-                ) ENGINE=InnoDB"
-            ))
-            .await
-            .map_err(into_error)?;
-        }
-
-        conn.query_drop(format!(
-            "CREATE TABLE IF NOT EXISTS {} (
-                k VARBINARY(255) NOT NULL,
-                v LONGBLOB NOT NULL,
-                PRIMARY KEY (k)
-            ) ENGINE=InnoDB",
-            char::from(SUBSPACE_BLOBS),
-        ))
-        .await
-        .map_err(into_error)?;
-
-        for table in [SUBSPACE_INDEXES, SUBSPACE_REGISTRY_IDX] {
-            let table = char::from(table);
-            conn.query_drop(format!(
-                "CREATE TABLE IF NOT EXISTS {table} (
-                    k BLOB,
-                    PRIMARY KEY (k(400))
-                ) ENGINE=InnoDB"
-            ))
-            .await
-            .map_err(into_error)?;
-        }
-
-        for table in [SUBSPACE_COUNTER, SUBSPACE_QUOTA, SUBSPACE_IN_MEMORY_COUNTER] {
             conn.query_drop(format!(
                 "CREATE TABLE IF NOT EXISTS {} (
-                k VARBINARY(255) NOT NULL,
+                    k VARBINARY(255) NOT NULL,
-                v BIGINT NOT NULL DEFAULT 0,
+                    v LONGBLOB NOT NULL,
-                PRIMARY KEY (k)
+                    PRIMARY KEY (k)
-            ) ENGINE=InnoDB",
+                ) ENGINE=InnoDB",
-                char::from(table)
+                char::from(SUBSPACE_BLOBS),
             ))
             .await
             .map_err(into_error)?;
-        }
 
-        Ok(())
+            for table in [SUBSPACE_INDEXES, SUBSPACE_REGISTRY_IDX] {
+                let table = char::from(table);
+                conn.query_drop(format!(
+                    "CREATE TABLE IF NOT EXISTS {table} (
+                        k BLOB,
+                        PRIMARY KEY (k(400))
+                    ) ENGINE=InnoDB"
+                ))
+                .await
+                .map_err(into_error)?;
+            }
+
+            for table in [SUBSPACE_COUNTER, SUBSPACE_QUOTA, SUBSPACE_IN_MEMORY_COUNTER] {
+                conn.query_drop(format!(
+                    "CREATE TABLE IF NOT EXISTS {} (
+                    k VARBINARY(255) NOT NULL,
+                    v BIGINT NOT NULL DEFAULT 0,
+                    PRIMARY KEY (k)
+                ) ENGINE=InnoDB",
+                    char::from(table)
+                ))
+                .await
+                .map_err(into_error)?;
+            }
+
+            Ok(())
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn create_search_tables(&self) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
+        let limit = self.timeouts.maintenance;
+        let result = tokio::time::timeout(limit, async {
+            create_search_tables::<EmailSearchField>(&mut conn).await?;
+            create_search_tables::<CalendarSearchField>(&mut conn).await?;
+            create_search_tables::<ContactSearchField>(&mut conn).await?;
+            //create_search_tables::<FileSearchField>(&mut conn).await?;
+            create_search_tables::<TracingSearchField>(&mut conn).await?;
 
-        create_search_tables::<EmailSearchField>(&mut conn).await?;
+            Ok(())
-        create_search_tables::<CalendarSearchField>(&mut conn).await?;
+        })
-        create_search_tables::<ContactSearchField>(&mut conn).await?;
+        .await;
-        //create_search_tables::<FileSearchField>(&mut conn).await?;
+        bounded(conn, result, limit)
-        create_search_tables::<TracingSearchField>(&mut conn).await?;
-
-        Ok(())
     }
 }
 

crates/store/src/backend/mysql/mod.rs

@@ -6,6 +6,7 @@
  * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
+use crate::backend::query_timeout::QueryTimeouts;
 use crate::{
     search::{
         CalendarSearchField, ContactSearchField, EmailSearchField, FileSearchField, SearchField,
@@ -14,7 +15,7 @@ use crate::{
     write::SearchIndex,
 };
 use mysql_async::Pool;
-use std::fmt::Display;
+use std::{fmt::Display, time::Duration};
 
 pub mod blob;
 pub mod lookup;
@@ -25,6 +26,72 @@ pub mod write;
 
 pub struct MysqlStore {
     pub(crate) conn_pool: Pool,
+    /// inbuxa: client-side query limits (see backend::query_timeout)
+    pub(crate) timeouts: QueryTimeouts,
+}
+
+/// inbuxa: how long a request waits for a pooled connection (including
+/// opening one). mysql_async's pool has no wait timeout, so upstream waited
+/// forever when the server stopped answering.
+pub(crate) const POOL_WAIT_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(30);
+/// inbuxa: idle time before TCP keepalive probes start.
+pub(crate) const POOL_KEEPALIVE_IDLE: std::time::Duration = std::time::Duration::from_secs(60);
+
+impl MysqlStore {
+    /// inbuxa: a pooled connection, or an error once POOL_WAIT_TIMEOUT has
+    /// passed without one.
+    pub(crate) async fn conn(&self) -> trc::Result<mysql_async::Conn> {
+        pool_conn(&self.conn_pool, POOL_WAIT_TIMEOUT).await
+    }
+}
+
+pub(crate) async fn pool_conn(
+    pool: &Pool,
+    wait: std::time::Duration,
+) -> trc::Result<mysql_async::Conn> {
+    match tokio::time::timeout(wait, pool.get_conn()).await {
+        Ok(result) => result.map_err(into_error),
+        Err(_) => Err(trc::StoreEvent::MysqlError
+            .reason("Timed out waiting for a database connection")
+            .details(format!("No connection within {} s", wait.as_secs()))),
+    }
+}
+
+/// inbuxa: the error for an operation that ran past its time limit.
+pub(crate) fn query_timeout_error(limit: Duration) -> trc::Error {
+    trc::StoreEvent::MysqlError
+        .reason("Query timed out")
+        .details(format!(
+            "No answer from the database within {} s",
+            limit.as_secs()
+        ))
+}
+
+/// inbuxa: ends an operation run on `conn` under `limit`. When it ran out,
+/// the connection is closed rather than returned to the pool: a query may
+/// still be in flight on it, or a transaction open. Conn::disconnect marks
+/// the connection closed before it sends anything, so even when the server
+/// doesn't answer and the attempt is dropped, the pool discards it instead
+/// of waiting to clean it up.
+pub(crate) fn bounded<T>(
+    conn: mysql_async::Conn,
+    result: Result<trc::Result<T>, tokio::time::error::Elapsed>,
+    limit: Duration,
+) -> trc::Result<T> {
+    match result {
+        Ok(result) => result,
+        Err(_) => {
+            discard(conn);
+            Err(query_timeout_error(limit))
+        }
+    }
+}
+
+/// inbuxa: closes a connection whose state is unknown (see bounded).
+pub(crate) fn discard(conn: mysql_async::Conn) {
+    tokio::spawn(async move {
+        let _ = tokio::time::timeout(Duration::from_secs(1), conn.disconnect()).await;
+    });
 }
 
 #[inline(always)]

crates/store/src/backend/mysql/read.rs

@@ -2,9 +2,11 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use super::{MysqlStore, into_error, is_timeout_error};
+use super::{MysqlStore, bounded, discard, into_error, is_timeout_error, query_timeout_error};
 use crate::{Deserialize, IterateParams, Key, ValueKey, write::ValueClass};
 use futures::TryStreamExt;
 use mysql_async::{Row, prelude::Queryable};
@@ -14,41 +16,51 @@ impl MysqlStore {
     where
         U: Deserialize + 'static,
     {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prep(format!(
+        let result = tokio::time::timeout(limit, async {
-                "SELECT v FROM {} WHERE k = ?",
+            let s = conn
-                char::from(key.subspace())
+                .prep(format!(
-            ))
+                    "SELECT v FROM {} WHERE k = ?",
-            .await
+                    char::from(key.subspace())
-            .map_err(into_error)?;
+                ))
-        let key = key.serialize(0);
+                .await
-        conn.exec_first::<Vec<u8>, _, _>(&s, (&key,))
+                .map_err(into_error)?;
-            .await
+            let key = key.serialize(0);
-            .map_err(into_error)
+            conn.exec_first::<Vec<u8>, _, _>(&s, (&key,))
-            .and_then(|r| {
+                .await
-                if let Some(r) = r {
+                .map_err(into_error)
-                    Ok(Some(U::deserialize_owned_with_key(&key, r)?))
+                .and_then(|r| {
-                } else {
+                    if let Some(r) = r {
-                    Ok(None)
+                        Ok(Some(U::deserialize_owned_with_key(&key, r)?))
-                }
+                    } else {
-            })
+                        Ok(None)
+                    }
+                })
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn key_exists(&self, key: impl Key) -> trc::Result<bool> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prep(format!(
+        let result = tokio::time::timeout(limit, async {
-                "SELECT 1 FROM {} WHERE k = ?",
+            let s = conn
-                char::from(key.subspace())
+                .prep(format!(
-            ))
+                    "SELECT 1 FROM {} WHERE k = ?",
-            .await
+                    char::from(key.subspace())
-            .map_err(into_error)?;
+                ))
-        let key = key.serialize(0);
+                .await
-        conn.exec_first::<u8, _, _>(&s, (&key,))
+                .map_err(into_error)?;
-            .await
+            let key = key.serialize(0);
-            .map_err(into_error)
+            conn.exec_first::<u8, _, _>(&s, (&key,))
-            .map(|r| r.is_some())
+                .await
+                .map_err(into_error)
+                .map(|r| r.is_some())
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn iterate<T: Key>(
@@ -56,34 +68,42 @@ impl MysqlStore {
         params: IterateParams<T>,
         mut cb: impl for<'x> FnMut(&'x [u8], &'x [u8]) -> trc::Result<bool> + Sync + Send,
     ) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
         let table = char::from(params.begin.subspace());
         let begin = params.begin.serialize(0);
         let end = params.end.serialize(0);
         let keys = if params.values { "k, v" } else { "k" };
 
-        let s = conn
+        // inbuxa: a scan may run for hours, so the query limit bounds each
-            .prep(&match (params.first, params.ascending) {
+        // wait for the database (preparing, the query starting, the next
-                (true, true) => {
+        // row) rather than the scan. A wait that runs out closes the
-                    format!(
+        // connection.
-                        "SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k ASC LIMIT 1"
+        let limit = self.timeouts.query;
-                    )
+        let query = match (params.first, params.ascending) {
-                }
+            (true, true) => {
-                (true, false) => {
+                format!("SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k ASC LIMIT 1")
-                    format!(
+            }
-                        "SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k DESC LIMIT 1"
+            (true, false) => {
-                    )
+                format!(
-                }
+                    "SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k DESC LIMIT 1"
-                (false, true) => {
+                )
-                    format!("SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k ASC")
+            }
-                }
+            (false, true) => {
-                (false, false) => {
+                format!("SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k ASC")
-                    format!("SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k DESC")
+            }
-                }
+            (false, false) => {
-            })
+                format!("SELECT {keys} FROM {table} WHERE k >= ? AND k <= ? ORDER BY k DESC")
-            .await
+            }
-            .map_err(into_error)?;
+        };
+        let s = match tokio::time::timeout(limit, conn.prep(&query)).await {
+            Ok(s) => s.map_err(into_error)?,
+            Err(_) => {
+                discard(conn);
+                return Err(query_timeout_error(limit));
+            }
+        };
         let mut from = begin;
+        let mut stalled = false;
         let mut to = end;
         let mut resume_key = None;
 
@@ -92,13 +112,26 @@ impl MysqlStore {
             let mut timed_out = false;
 
             {
-                let mut rows = conn
+                let mut rows = match tokio::time::timeout(
-                    .exec_stream::<Row, _, _>(&s, (from.clone(), to.clone()))
+                    limit,
-                    .await
+                    conn.exec_stream::<Row, _, _>(&s, (from.clone(), to.clone())),
-                    .map_err(into_error)?;
+                )
+                .await
+                {
+                    Ok(rows) => rows.map_err(into_error)?,
+                    // Leaves the scan loop for the timeout below
+                    Err(_) => break,
+                };
 
                 loop {
-                    match rows.try_next().await {
+                    let next = match tokio::time::timeout(limit, rows.try_next()).await {
+                        Ok(next) => next,
+                        Err(_) => {
+                            stalled = true;
+                            break;
+                        }
+                    };
+                    match next {
                         Ok(Some(mut row)) => {
                             let value = if params.values {
                                 row.take_opt::<Vec<u8>, _>(1)
@@ -134,6 +167,10 @@ impl MysqlStore {
                 }
             }
 
+            if stalled {
+                break;
+            }
+
             match last_key {
                 Some(last_key) if timed_out => {
                     if params.ascending {
@@ -146,6 +183,9 @@ impl MysqlStore {
                 _ => return Ok(()),
             }
         }
+
+        discard(conn);
+        Err(query_timeout_error(limit))
     }
 
     pub(crate) async fn get_counter(
@@ -155,15 +195,20 @@ impl MysqlStore {
         let key = key.into();
         let table = char::from(key.subspace());
         let key = key.serialize(0);
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prep(format!("SELECT v FROM {table} WHERE k = ?"))
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prep(format!("SELECT v FROM {table} WHERE k = ?"))
-        match conn.exec_first::<i64, _, _>(&s, (key,)).await {
+                .await
-            Ok(Some(num)) => Ok(num),
+                .map_err(into_error)?;
-            Ok(None) => Ok(0),
+            match conn.exec_first::<i64, _, _>(&s, (key,)).await {
-            Err(e) => Err(into_error(e)),
+                Ok(Some(num)) => Ok(num),
-        }
+                Ok(None) => Ok(0),
+                Err(e) => Err(into_error(e)),
+            }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }

crates/store/src/backend/mysql/search.rs

@@ -10,8 +10,8 @@ use crate::{
     backend::{
         MAX_TOKEN_LENGTH,
         mysql::{
-            DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, MysqlSearchField, MysqlStore, into_error,
+            DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, MysqlSearchField, MysqlStore, bounded,
-            is_timeout_error,
+            into_error, is_timeout_error,
         },
     },
     search::{
@@ -26,58 +26,63 @@ use std::fmt::Write;
 
 impl MysqlStore {
     pub async fn index(&self, documents: Vec<IndexDocument>) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let mut tx_opts = TxOpts::default();
+        let limit = self.timeouts.query;
-        tx_opts
+        let result = tokio::time::timeout(limit, async {
-            .with_consistent_snapshot(false)
+            let mut tx_opts = TxOpts::default();
-            .with_isolation_level(IsolationLevel::ReadCommitted);
+            tx_opts
-        let mut trx = conn.start_transaction(tx_opts).await.map_err(into_error)?;
+                .with_consistent_snapshot(false)
+                .with_isolation_level(IsolationLevel::ReadCommitted);
+            let mut trx = conn.start_transaction(tx_opts).await.map_err(into_error)?;
 
-        for document in documents {
+            for document in documents {
-            let index = document.index;
+                let index = document.index;
-            let primary_keys = index.primary_keys();
+                let primary_keys = index.primary_keys();
-            let all_fields = index.all_fields();
+                let all_fields = index.all_fields();
-            let mut fields = document.fields;
+                let mut fields = document.fields;
-            let mut values = Vec::with_capacity(fields.len() + 2);
+                let mut values = Vec::with_capacity(fields.len() + 2);
-            let mut query = format!("INSERT INTO {} (", index.mysql_table());
+                let mut query = format!("INSERT INTO {} (", index.mysql_table());
 
-            for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
+                for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
-                if i > 0 {
+                    if i > 0 {
-                    query.push(',');
+                        query.push(',');
+                    }
+                    query.push_str(field.column());
                 }
-                query.push_str(field.column());
+
+                query.push_str(") VALUES (");
+
+                for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
+                    if i > 0 {
+                        query.push(',');
+                    }
+
+                    if let Some(value) = fields.remove(field) {
+                        query.push('?');
+                        values.push(value);
+                    } else {
+                        query.push_str("NULL");
+                    }
+                }
+
+                query.push_str(") ON DUPLICATE KEY UPDATE ");
+                for (i, field) in all_fields.iter().enumerate() {
+                    if i > 0 {
+                        query.push(',');
+                    }
+                    let column = field.column();
+                    let _ = write!(&mut query, "{column} = VALUES({column})");
+                }
+
+                let s = trx.prep(&query).await.map_err(into_error)?;
+
+                trx.exec_drop(&s, values).await.map_err(into_error)?;
             }
 
-            query.push_str(") VALUES (");
+            trx.commit().await.map_err(into_error)
-
+        })
-            for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
+        .await;
-                if i > 0 {
+        bounded(conn, result, limit)
-                    query.push(',');
-                }
-
-                if let Some(value) = fields.remove(field) {
-                    query.push('?');
-                    values.push(value);
-                } else {
-                    query.push_str("NULL");
-                }
-            }
-
-            query.push_str(") ON DUPLICATE KEY UPDATE ");
-            for (i, field) in all_fields.iter().enumerate() {
-                if i > 0 {
-                    query.push(',');
-                }
-                let column = field.column();
-                let _ = write!(&mut query, "{column} = VALUES({column})");
-            }
-
-            let s = trx.prep(&query).await.map_err(into_error)?;
-
-            trx.exec_drop(&s, values).await.map_err(into_error)?;
-        }
-
-        trx.commit().await.map_err(into_error)
     }
 
     pub async fn query<R: SearchDocumentId>(
@@ -96,13 +101,18 @@ impl MysqlStore {
             build_sort(&mut query, sort);
         }
 
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn.prep(query).await.map_err(into_error)?;
+        let limit = self.timeouts.query;
+        let result = tokio::time::timeout(limit, async {
+            let s = conn.prep(query).await.map_err(into_error)?;
 
-        conn.exec::<i64, _, _>(s, params)
+            conn.exec::<i64, _, _>(s, params)
-            .await
+                .await
-            .map(|r| r.into_iter().map(|r| R::from_u64(r as u64)).collect())
+                .map(|r| r.into_iter().map(|r| R::from_u64(r as u64)).collect())
-            .map_err(into_error)
+                .map_err(into_error)
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub async fn unindex(&self, filter: SearchQuery) -> trc::Result<u64> {
@@ -110,41 +120,48 @@ impl MysqlStore {
         let mut query = format!("DELETE FROM {table} ");
         let params = build_filter(&mut query, &filter.filters);
 
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let s = conn.prep(&query).await.map_err(into_error)?;
+        let limit = self.timeouts.maintenance;
+        let result = tokio::time::timeout(limit, async {
+            let s = conn.prep(&query).await.map_err(into_error)?;
 
-        match conn.exec_drop(s, params.clone()).await {
+            match conn.exec_drop(s, params.clone()).await {
-            Ok(_) => return Ok(conn.affected_rows()),
+                Ok(_) => return Ok(conn.affected_rows()),
-            Err(err) if is_timeout_error(&err) => (),
+                Err(err) if is_timeout_error(&err) => (),
-            Err(err) => return Err(into_error(err)),
+                Err(err) => return Err(into_error(err)),
-        }
+            }
 
-        let mut chunk_size = DELETE_CHUNK_SIZE;
+            let mut chunk_size = DELETE_CHUNK_SIZE;
-        let mut deleted = 0;
+            let mut deleted = 0;
-
-        loop {
-            let s = conn
-                .prep(format!("{query} LIMIT {chunk_size}"))
-                .await
-                .map_err(into_error)?;
 
             loop {
-                match conn.exec_drop(&s, params.clone()).await {
+                let s = conn
-                    Ok(_) => {
+                    .prep(format!("{query} LIMIT {chunk_size}"))
-                        let affected = conn.affected_rows();
+                    .await
-                        if affected == 0 {
+                    .map_err(into_error)?;
-                            return Ok(deleted);
+
+                loop {
+                    match conn.exec_drop(&s, params.clone()).await {
+                        Ok(_) => {
+                            let affected = conn.affected_rows();
+                            if affected == 0 {
+                                return Ok(deleted);
+                            }
+                            deleted += affected;
                         }
-                        deleted += affected;
+                        Err(err)
+                            if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE =>
+                        {
+                            chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                            break;
+                        }
+                        Err(err) => return Err(into_error(err)),
                     }
-                    Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
-                        chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
-                        break;
-                    }
-                    Err(err) => return Err(into_error(err)),
                 }
             }
-        }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }
 

crates/store/src/backend/mysql/write.rs

@@ -2,9 +2,13 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use super::{DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, MysqlStore, into_error, is_timeout_error};
+use super::{
+    DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, MysqlStore, bounded, into_error, is_timeout_error,
+};
 use crate::{
     IndexKey, Key, LogKey, SUBSPACE_COUNTER, SUBSPACE_IN_MEMORY_COUNTER, SUBSPACE_QUOTA,
     SUBSPACE_REGISTRY_IDX,
@@ -29,42 +33,46 @@ impl MysqlStore {
     pub(crate) async fn write(&self, mut batch: Batch<'_>) -> trc::Result<AssignedIds> {
         let start = Instant::now();
         let mut retry_count = 0;
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-
+        let limit = self.timeouts.query;
-        loop {
+        let result = tokio::time::timeout(limit, async {
-            let err = match self.write_trx(&mut conn, &mut batch).await {
+            loop {
-                Ok(result) => {
+                let err = match self.write_trx(&mut conn, &mut batch).await {
-                    return Ok(result);
+                    Ok(result) => {
-                }
+                        return Ok(result);
-                Err(err) => err,
-            };
-
-            let _ = conn.query_drop("ROLLBACK;").await;
-
-            match err {
-                CommitError::Mysql(Error::Server(err))
-                    if [1062, 1213].contains(&err.code)
-                        && retry_count < MAX_COMMIT_ATTEMPTS
-                        && start.elapsed() < MAX_COMMIT_TIME => {}
-                /*CommitError::Retry => {
-                    if retry_count > MAX_COMMIT_ATTEMPTS || start.elapsed() > MAX_COMMIT_TIME {
-                        return Err(trc::StoreEvent::AssertValueFailed
-                            .into_err()
-                            .caused_by(trc::location!()));
                     }
-                }*/
+                    Err(err) => err,
-                CommitError::Mysql(err) => {
+                };
-                    return Err(into_error(err));
-                }
-                CommitError::Internal(err) => {
-                    return Err(err);
-                }
-            }
 
-            let backoff = rand::rng().random_range(50..=300);
+                let _ = conn.query_drop("ROLLBACK;").await;
-            tokio::time::sleep(Duration::from_millis(backoff)).await;
+
-            retry_count += 1;
+                match err {
-        }
+                    CommitError::Mysql(Error::Server(err))
+                        if [1062, 1213].contains(&err.code)
+                            && retry_count < MAX_COMMIT_ATTEMPTS
+                            && start.elapsed() < MAX_COMMIT_TIME => {}
+                    /*CommitError::Retry => {
+                        if retry_count > MAX_COMMIT_ATTEMPTS || start.elapsed() > MAX_COMMIT_TIME {
+                            return Err(trc::StoreEvent::AssertValueFailed
+                                .into_err()
+                                .caused_by(trc::location!()));
+                        }
+                    }*/
+                    CommitError::Mysql(err) => {
+                        return Err(into_error(err));
+                    }
+                    CommitError::Internal(err) => {
+                        return Err(err);
+                    }
+                }
+
+                let backoff = rand::rng().random_range(50..=300);
+                tokio::time::sleep(Duration::from_millis(backoff)).await;
+                retry_count += 1;
+            }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     async fn write_trx(
@@ -382,72 +390,82 @@ impl MysqlStore {
     }
 
     pub(crate) async fn purge_store(&self) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        for subspace in [SUBSPACE_QUOTA, SUBSPACE_COUNTER, SUBSPACE_IN_MEMORY_COUNTER] {
+        let limit = self.timeouts.maintenance;
-            purge_table(&mut conn, char::from(subspace)).await?;
+        let result = tokio::time::timeout(limit, async {
-        }
+            for subspace in [SUBSPACE_QUOTA, SUBSPACE_COUNTER, SUBSPACE_IN_MEMORY_COUNTER] {
+                purge_table(&mut conn, char::from(subspace)).await?;
+            }
 
-        Ok(())
+            Ok(())
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn delete_range(&self, from: impl Key, to: impl Key) -> trc::Result<()> {
-        let mut conn = self.conn_pool.get_conn().await.map_err(into_error)?;
+        let mut conn = self.conn().await?;
-        let table = char::from(from.subspace());
+        let limit = self.timeouts.maintenance;
-        let mut from = from.serialize(0);
+        let result = tokio::time::timeout(limit, async {
-        let to = to.serialize(0);
+            let table = char::from(from.subspace());
+            let mut from = from.serialize(0);
+            let to = to.serialize(0);
 
-        let delete = conn
+            let delete = conn
-            .prep(format!("DELETE FROM {table} WHERE k >= ? AND k < ?"))
+                .prep(format!("DELETE FROM {table} WHERE k >= ? AND k < ?"))
-            .await
-            .map_err(into_error)?;
-
-        match conn.exec_drop(&delete, (&from, &to)).await {
-            Ok(_) => return Ok(()),
-            Err(err) if is_timeout_error(&err) => (),
-            Err(err) => return Err(into_error(err)),
-        }
-
-        let mut chunk_size = DELETE_CHUNK_SIZE;
-
-        loop {
-            let boundary = conn
-                .prep(format!(
-                    "SELECT k FROM {table} WHERE k >= ? AND k < ? ORDER BY k ASC LIMIT 1 OFFSET {chunk_size}"
-                ))
                 .await
                 .map_err(into_error)?;
 
+            match conn.exec_drop(&delete, (&from, &to)).await {
+                Ok(_) => return Ok(()),
+                Err(err) if is_timeout_error(&err) => (),
+                Err(err) => return Err(into_error(err)),
+            }
+
+            let mut chunk_size = DELETE_CHUNK_SIZE;
+
             loop {
-                let next = match conn
+                let boundary = conn
-                    .exec_first::<Vec<u8>, _, _>(&boundary, (&from, &to))
+                    .prep(format!(
+                        "SELECT k FROM {table} WHERE k >= ? AND k < ? ORDER BY k ASC LIMIT 1 OFFSET {chunk_size}"
+                    ))
                     .await
-                {
+                    .map_err(into_error)?;
-                    Ok(next) => next,
-                    Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
-                        chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
-                        break;
-                    }
-                    Err(err) => return Err(into_error(err)),
-                };
 
-                match conn
+                loop {
-                    .exec_drop(&delete, (&from, next.as_ref().unwrap_or(&to)))
+                    let next = match conn
-                    .await
+                        .exec_first::<Vec<u8>, _, _>(&boundary, (&from, &to))
-                {
+                        .await
-                    Ok(_) => (),
+                    {
-                    Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
+                        Ok(next) => next,
-                        chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                        Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
-                        break;
+                            chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
-                    }
+                            break;
-                    Err(err) => return Err(into_error(err)),
+                        }
-                }
+                        Err(err) => return Err(into_error(err)),
+                    };
 
-                match next {
+                    match conn
-                    Some(next) => from = next,
+                        .exec_drop(&delete, (&from, next.as_ref().unwrap_or(&to)))
-                    None => return Ok(()),
+                        .await
+                    {
+                        Ok(_) => (),
+                        Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
+                            chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                            break;
+                        }
+                        Err(err) => return Err(into_error(err)),
+                    }
+
+                    match next {
+                        Some(next) => from = next,
+                        None => return Ok(()),
+                    }
                 }
             }
-        }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }
 

crates/store/src/backend/postgres/blob.rs

@@ -2,13 +2,15 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use std::ops::Range;
 
 use crate::backend::postgres::into_pool_error;
 
-use super::{PostgresStore, into_error};
+use super::{PostgresStore, bounded, into_error};
 
 impl PostgresStore {
     pub(crate) async fn get_blob(
@@ -17,53 +19,68 @@ impl PostgresStore {
         range: Range<usize>,
     ) -> trc::Result<Option<Vec<u8>>> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prepare_cached("SELECT v FROM t WHERE k = $1")
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prepare_cached("SELECT v FROM t WHERE k = $1")
-        conn.query_opt(&s, &[&key])
+                .await
-            .await
+                .map_err(into_error)?;
-            .and_then(|row| {
+            conn.query_opt(&s, &[&key])
-                if let Some(row) = row {
+                .await
-                    Ok(Some(if range.start == 0 && range.end == usize::MAX {
+                .and_then(|row| {
-                        row.try_get::<_, Vec<u8>>(0)?
+                    if let Some(row) = row {
+                        Ok(Some(if range.start == 0 && range.end == usize::MAX {
+                            row.try_get::<_, Vec<u8>>(0)?
+                        } else {
+                            let bytes = row.try_get::<_, &[u8]>(0)?;
+                            bytes
+                                .get(range.start..std::cmp::min(bytes.len(), range.end))
+                                .unwrap_or_default()
+                                .to_vec()
+                        }))
                     } else {
-                        let bytes = row.try_get::<_, &[u8]>(0)?;
+                        Ok(None)
-                        bytes
+                    }
-                            .get(range.start..std::cmp::min(bytes.len(), range.end))
+                })
-                            .unwrap_or_default()
+                .map_err(into_error)
-                            .to_vec()
+        })
-                    }))
+        .await;
-                } else {
+        bounded(conn, result, limit)
-                    Ok(None)
-                }
-            })
-            .map_err(into_error)
     }
 
     pub(crate) async fn put_blob(&self, key: &[u8], data: &[u8]) -> trc::Result<()> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prepare_cached(
+        let result = tokio::time::timeout(limit, async {
-                "INSERT INTO t (k, v) VALUES ($1, $2) ON CONFLICT (k) DO UPDATE SET v = EXCLUDED.v",
+            let s = conn
-            )
+                .prepare_cached(
-            .await
+                    "INSERT INTO t (k, v) VALUES ($1, $2) ON CONFLICT (k) DO UPDATE SET v = EXCLUDED.v",
-            .map_err(into_error)?;
+                )
-        conn.execute(&s, &[&key, &data])
+                .await
-            .await
+                .map_err(into_error)?;
-            .map_err(into_error)
+            conn.execute(&s, &[&key, &data])
-            .map(|_| ())
+                .await
+                .map_err(into_error)
+                .map(|_| ())
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn delete_blob(&self, key: &[u8]) -> trc::Result<bool> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prepare_cached("DELETE FROM t WHERE k = $1")
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prepare_cached("DELETE FROM t WHERE k = $1")
-        conn.execute(&s, &[&key])
+                .await
-            .await
+                .map_err(into_error)?;
-            .map_err(into_error)
+            conn.execute(&s, &[&key])
-            .map(|hits| hits > 0)
+                .await
+                .map_err(into_error)
+                .map(|hits| hits > 0)
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }

crates/store/src/backend/postgres/lookup.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::{QueryResult, QueryType, backend::postgres::into_pool_error};
@@ -12,7 +14,7 @@ use tokio_postgres::types::{FromSql, ToSql, Type};
 
 use crate::IntoRows;
 
-use super::{PostgresStore, into_error};
+use super::{PostgresStore, bounded, into_error};
 
 impl PostgresStore {
     pub(crate) async fn sql_query<T: QueryResult>(
@@ -21,33 +23,38 @@ impl PostgresStore {
         params_: &[crate::Value<'_>],
     ) -> trc::Result<T> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn.prepare_cached(query).await.map_err(into_error)?;
+        let limit = self.timeouts.query;
-        let params = params_
+        let result = tokio::time::timeout(limit, async {
-            .iter()
+            let s = conn.prepare_cached(query).await.map_err(into_error)?;
-            .map(|v| v as &(dyn tokio_postgres::types::ToSql + Sync))
+            let params = params_
-            .collect::<Vec<_>>();
+                .iter()
+                .map(|v| v as &(dyn tokio_postgres::types::ToSql + Sync))
+                .collect::<Vec<_>>();
 
-        match T::query_type() {
+            match T::query_type() {
-            QueryType::Execute => conn
+                QueryType::Execute => conn
-                .execute(&s, params.as_slice())
+                    .execute(&s, params.as_slice())
-                .await
-                .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_exec(r as usize))),
-            QueryType::Exists => {
-                let rows = conn.query_raw(&s, params).await.map_err(into_error)?;
-                pin_mut!(rows);
-                rows.try_next()
                     .await
-                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_exists(r.is_some())))
+                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_exec(r as usize))),
+                QueryType::Exists => {
+                    let rows = conn.query_raw(&s, params).await.map_err(into_error)?;
+                    pin_mut!(rows);
+                    rows.try_next()
+                        .await
+                        .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_exists(r.is_some())))
+                }
+                QueryType::QueryOne => conn
+                    .query_opt(&s, params.as_slice())
+                    .await
+                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_one(r))),
+                QueryType::QueryAll => conn
+                    .query(&s, params.as_slice())
+                    .await
+                    .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_all(r))),
             }
-            QueryType::QueryOne => conn
+        })
-                .query_opt(&s, params.as_slice())
+        .await;
-                .await
+        bounded(conn, result, limit)
-                .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_one(r))),
-            QueryType::QueryAll => conn
-                .query(&s, params.as_slice())
-                .await
-                .map_or_else(|e| Err(into_error(e)), |r| Ok(T::from_query_all(r))),
-        }
     }
 }
 

crates/store/src/backend/postgres/main.rs

@@ -6,7 +6,7 @@
  * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use super::{PostgresStore, into_error};
+use super::{PostgresStore, bounded, into_error};
 use crate::{
     backend::postgres::{
         PsqlSearchField, into_pool_error,
@@ -22,11 +22,34 @@ use crate::{
 use ::registry::schema::{enums::PostgreSqlRecyclingMethod, structs};
 use ahash::AHashSet;
 use deadpool_postgres::{
-    Config, ManagerConfig, Object, Pool, PoolConfig, RecyclingMethod, Runtime,
+    Config, ManagerConfig, Object, Pool, PoolConfig, RecyclingMethod, Runtime, Timeouts,
 };
+use std::time::Duration;
 use tokio_postgres::NoTls;
 use utils::tls::rustls_client_config;
 
+/// inbuxa: how long a request waits for a pooled connection.
+pub(crate) const POOL_WAIT_TIMEOUT: Duration = Duration::from_secs(30);
+/// inbuxa: how long opening a connection may take when the store sets no
+/// timeout of its own.
+pub(crate) const POOL_CREATE_TIMEOUT: Duration = Duration::from_secs(15);
+/// inbuxa: how long checking a pooled connection before reuse may take.
+pub(crate) const POOL_RECYCLE_TIMEOUT: Duration = Duration::from_secs(10);
+/// inbuxa: idle time before TCP keepalive probes start.
+pub(crate) const POOL_KEEPALIVE_IDLE: Duration = Duration::from_secs(60);
+
+/// inbuxa: the pool's timeouts. Opening a connection is bounded by the
+/// store's own timeout when it has one; waiting for one covers at least that
+/// long, so a slow connect isn't cut short by the wait.
+pub(crate) fn pool_timeouts(connect_timeout: Option<Duration>) -> Timeouts {
+    let create = connect_timeout.unwrap_or(POOL_CREATE_TIMEOUT);
+    Timeouts {
+        wait: POOL_WAIT_TIMEOUT.max(create).into(),
+        create: create.into(),
+        recycle: POOL_RECYCLE_TIMEOUT.into(),
+    }
+}
+
 impl PostgresStore {
     pub async fn open(config: structs::PostgreSqlStore) -> Result<Store, String> {
         // inbuxa: ST-15: where the primary is, to tell a replica from it
@@ -46,9 +69,20 @@ impl PostgresStore {
                 PostgreSqlRecyclingMethod::Clean => RecyclingMethod::Clean,
             },
         });
-        if let Some(max_conn) = config.pool_max_connections {
+        // inbuxa: upstream set no pool timeouts, so a request waited for a
-            cfg.pool = PoolConfig::new(max_conn as usize).into();
+        // free connection, or for one to be made or recycled, for as long as
-        }
+        // it took: forever when the server stopped answering. A worker now
+        // gets an error instead and the task or request is retried.
+        let mut pool = config
+            .pool_max_connections
+            .map(|max_conn| PoolConfig::new(max_conn as usize))
+            .unwrap_or_default();
+        pool.timeouts = pool_timeouts(cfg.connect_timeout);
+        cfg.pool = pool.into();
+        // Notice a server that went away without closing the connection in
+        // minutes rather than the system default of two hours
+        cfg.keepalives = true.into();
+        cfg.keepalives_idle = POOL_KEEPALIVE_IDLE.into();
 
         let primary_pool = if config.use_tls {
             cfg.create_pool(
@@ -85,6 +119,7 @@ impl PostgresStore {
                 Store::PostgreSQL(Arc::new(PostgresStore {
                     conn_pool: pool,
                     ts_configs: ts_configs.clone(),
+                    timeouts: Default::default(),
                 })),
                 replica.host,
                 replica.port as u16,
@@ -95,6 +130,7 @@ impl PostgresStore {
         let primary = Store::PostgreSQL(Arc::new(PostgresStore {
             conn_pool: primary_pool,
             ts_configs,
+            timeouts: Default::default(),
         }));
 
         // ST-1: no replicas, no change
@@ -113,84 +149,92 @@ impl PostgresStore {
 
     pub(crate) async fn create_storage_tables(&self) -> trc::Result<()> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
+        let limit = self.timeouts.maintenance;
+        let result = tokio::time::timeout(limit, async {
+            for table in [
+                SUBSPACE_ACL,
+                SUBSPACE_TASK_QUEUE,
+                SUBSPACE_DELETED_ITEMS,
+                SUBSPACE_SPAM_SAMPLES,
+                crate::SUBSPACE_INBUXA, // inbuxa: masked email
+                SUBSPACE_BLOB_LINK,
+                SUBSPACE_IN_MEMORY_VALUE,
+                SUBSPACE_PROPERTY,
+                SUBSPACE_REGISTRY,
+                SUBSPACE_REGISTRY_PK,
+                SUBSPACE_QUEUE_MESSAGE,
+                SUBSPACE_QUEUE_EVENT,
+                SUBSPACE_REPORT_OUT,
+                SUBSPACE_REPORT_IN,
+                SUBSPACE_LOGS,
+                SUBSPACE_BLOBS,
+                SUBSPACE_DIRECTORY,
+                SUBSPACE_TELEMETRY_SPAN,
+                SUBSPACE_TELEMETRY_METRIC,
+            ] {
+                let table = char::from(table);
+                conn.execute(
+                    &format!(
+                        "CREATE TABLE IF NOT EXISTS {table} (
+                            k BYTEA PRIMARY KEY,
+                            v BYTEA NOT NULL
+                        )"
+                    ),
+                    &[],
+                )
+                .await
+                .map_err(into_error)?;
+            }
 
-        for table in [
+            for table in [SUBSPACE_INDEXES, SUBSPACE_REGISTRY_IDX] {
-            SUBSPACE_ACL,
+                let table = char::from(table);
-            SUBSPACE_TASK_QUEUE,
+                conn.execute(
-            SUBSPACE_DELETED_ITEMS,
+                    &format!(
-            SUBSPACE_SPAM_SAMPLES,
+                        "CREATE TABLE IF NOT EXISTS {table} (
-            crate::SUBSPACE_INBUXA, // inbuxa: masked email
+                            k BYTEA PRIMARY KEY
-            SUBSPACE_BLOB_LINK,
+                        )"
-            SUBSPACE_IN_MEMORY_VALUE,
+                    ),
-            SUBSPACE_PROPERTY,
+                    &[],
-            SUBSPACE_REGISTRY,
+                )
-            SUBSPACE_REGISTRY_PK,
+                .await
-            SUBSPACE_QUEUE_MESSAGE,
+                .map_err(into_error)?;
-            SUBSPACE_QUEUE_EVENT,
+            }
-            SUBSPACE_REPORT_OUT,
+
-            SUBSPACE_REPORT_IN,
+            for table in [SUBSPACE_COUNTER, SUBSPACE_QUOTA, SUBSPACE_IN_MEMORY_COUNTER] {
-            SUBSPACE_LOGS,
+                conn.execute(
-            SUBSPACE_BLOBS,
+                    &format!(
-            SUBSPACE_DIRECTORY,
+                        "CREATE TABLE IF NOT EXISTS {} (
-            SUBSPACE_TELEMETRY_SPAN,
-            SUBSPACE_TELEMETRY_METRIC,
-        ] {
-            let table = char::from(table);
-            conn.execute(
-                &format!(
-                    "CREATE TABLE IF NOT EXISTS {table} (
                         k BYTEA PRIMARY KEY,
-                        v BYTEA NOT NULL
+                        v BIGINT NOT NULL DEFAULT 0
-                    )"
+                    )",
-                ),
+                        char::from(table)
-                &[],
+                    ),
-            )
+                    &[],
-            .await
+                )
-            .map_err(into_error)?;
+                .await
-        }
+                .map_err(into_error)?;
+            }
 
-        for table in [SUBSPACE_INDEXES, SUBSPACE_REGISTRY_IDX] {
+            Ok(())
-            let table = char::from(table);
+        })
-            conn.execute(
+        .await;
-                &format!(
+        bounded(conn, result, limit)
-                    "CREATE TABLE IF NOT EXISTS {table} (
-                        k BYTEA PRIMARY KEY
-                    )"
-                ),
-                &[],
-            )
-            .await
-            .map_err(into_error)?;
-        }
-
-        for table in [SUBSPACE_COUNTER, SUBSPACE_QUOTA, SUBSPACE_IN_MEMORY_COUNTER] {
-            conn.execute(
-                &format!(
-                    "CREATE TABLE IF NOT EXISTS {} (
-                    k BYTEA PRIMARY KEY,
-                    v BIGINT NOT NULL DEFAULT 0
-                )",
-                    char::from(table)
-                ),
-                &[],
-            )
-            .await
-            .map_err(into_error)?;
-        }
-
-        Ok(())
     }
 
     pub(crate) async fn create_search_tables(&self) -> trc::Result<()> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
+        let limit = self.timeouts.maintenance;
+        let result = tokio::time::timeout(limit, async {
+            create_search_tables::<EmailSearchField>(&conn).await?;
+            create_search_tables::<CalendarSearchField>(&conn).await?;
+            create_search_tables::<ContactSearchField>(&conn).await?;
+            //create_search_tables::<FileSearchField>(&conn).await?;
+            create_search_tables::<TracingSearchField>(&conn).await?;
 
-        create_search_tables::<EmailSearchField>(&conn).await?;
+            Ok(())
-        create_search_tables::<CalendarSearchField>(&conn).await?;
+        })
-        create_search_tables::<ContactSearchField>(&conn).await?;
+        .await;
-        //create_search_tables::<FileSearchField>(&conn).await?;
+        bounded(conn, result, limit)
-        create_search_tables::<TracingSearchField>(&conn).await?;
-
-        Ok(())
     }
 }
 
@@ -231,12 +275,21 @@ async fn create_search_tables<T: SearchableField + PsqlSearchField + 'static>(
     for field in T::all_fields() {
         if field.is_text() || field.is_json() {
             let column_name = field.column();
+            // inbuxa: with GIN's default fastupdate=on, new entries wait in
+            // an unindexed pending list that every search scans in full
+            // until a VACUUM (or 4 MB of backlog) merges it. On a mailbox
+            // taking steady mail that list never drains and searches slow
+            // from milliseconds to hundreds of them. Pay the index update
+            // at insert time instead.
+            let index_name = format!("gin_{table_name}_{column_name}");
             let create_index_query = format!(
-                "CREATE INDEX IF NOT EXISTS gin_{table_name}_{column_name} ON {table_name} USING GIN({column_name})",
+                "CREATE INDEX IF NOT EXISTS {index_name} ON {table_name} USING GIN({column_name}) WITH (fastupdate = off)",
             );
             conn.execute(&create_index_query, &[])
                 .await
                 .map_err(into_error)?;
+            // Indexes made before this change keep fastupdate=on
+            disable_gin_fastupdate(conn, &index_name).await;
         }
 
         if field.is_indexed() {
@@ -253,6 +306,69 @@ async fn create_search_tables<T: SearchableField + PsqlSearchField + 'static>(
     Ok(())
 }
 
+/// inbuxa: turns fastupdate off on a GIN index made with the default and
+/// merges the pending list it has built up. Idempotent: an index that already
+/// has the option is left alone, so this costs one catalog read per index at
+/// startup. A failure is logged and startup goes on, since search still works,
+/// only slower.
+async fn disable_gin_fastupdate(conn: &Object, index_name: &str) {
+    if let Err(err) = try_disable_gin_fastupdate(conn, index_name).await {
+        trc::event!(
+            Store(trc::StoreEvent::PostgresqlError),
+            Details = format!("Failed to turn off fastupdate on search index {index_name}"),
+            Reason = err.to_string(),
+        );
+    }
+}
+
+async fn try_disable_gin_fastupdate(conn: &Object, index_name: &str) -> trc::Result<()> {
+    let options = conn
+        .query_opt(
+            "SELECT COALESCE(reloptions, '{}')::text[] FROM pg_class WHERE oid = to_regclass($1)",
+            &[&index_name],
+        )
+        .await
+        .map_err(into_error)?
+        .map(|row| row.try_get::<_, Vec<String>>(0))
+        .transpose()
+        .map_err(into_error)?;
+    let Some(options) = options else {
+        return Ok(());
+    };
+    if gin_fastupdate_is_off(&options) {
+        return Ok(());
+    }
+    // SET (fastupdate) takes a SHARE UPDATE EXCLUSIVE lock, which doesn't
+    // block reads or writes. Turning it off stops new entries going to the
+    // pending list but doesn't flush the entries already there.
+    conn.execute(
+        &format!("ALTER INDEX {index_name} SET (fastupdate = off)"),
+        &[],
+    )
+    .await
+    .map_err(into_error)?;
+    conn.query_one(
+        "SELECT gin_clean_pending_list($1::text::regclass)",
+        &[&index_name],
+    )
+    .await
+    .map_err(into_error)?;
+    Ok(())
+}
+
+/// Whether a relation's reloptions turn GIN's fastupdate off.
+fn gin_fastupdate_is_off(options: &[String]) -> bool {
+    options.iter().any(|option| {
+        option.split_once('=').is_some_and(|(name, value)| {
+            name.trim().eq_ignore_ascii_case("fastupdate")
+                && matches!(
+                    value.trim().to_ascii_lowercase().as_str(),
+                    "off" | "false" | "no" | "0" | "f" | "n"
+                )
+        })
+    })
+}
+
 async fn discover_ts_configs(pool: &Pool) -> AHashSet<&'static str> {
     let mut ts_configs = AHashSet::from_iter([PG_FALLBACK_LANG, PG_UNSTEMMED_LANG]);
 

crates/store/src/backend/postgres/mod.rs

@@ -6,6 +6,7 @@
  * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
+use crate::backend::query_timeout::QueryTimeouts;
 use crate::{
     search::{
         CalendarSearchField, ContactSearchField, EmailSearchField, FileSearchField, SearchField,
@@ -14,7 +15,8 @@ use crate::{
     write::SearchIndex,
 };
 use ahash::AHashSet;
-use deadpool_postgres::Pool;
+use deadpool_postgres::{Object, Pool};
+use std::time::Duration;
 use tokio_postgres::error::SqlState;
 
 pub mod blob;
@@ -28,6 +30,8 @@ pub mod write;
 pub struct PostgresStore {
     pub(crate) conn_pool: Pool,
     pub(crate) ts_configs: AHashSet<&'static str>,
+    /// inbuxa: client-side query limits (see backend::query_timeout)
+    pub(crate) timeouts: QueryTimeouts,
 }
 
 #[inline(always)]
@@ -72,6 +76,34 @@ pub(crate) fn is_timeout_error(err: &tokio_postgres::Error) -> bool {
     })
 }
 
+/// inbuxa: the error for an operation that ran past its time limit.
+pub(crate) fn query_timeout_error(limit: Duration) -> trc::Error {
+    trc::StoreEvent::PostgresqlError
+        .reason("Query timed out")
+        .details(format!(
+            "No answer from the database within {} s",
+            limit.as_secs()
+        ))
+}
+
+/// inbuxa: ends an operation run on `conn` under `limit`. When it ran out,
+/// the connection is taken out of the pool and closed: a query may still be
+/// in flight on it, or a transaction open, so it can't be handed to the
+/// next caller.
+pub(crate) fn bounded<T>(
+    conn: Object,
+    result: Result<trc::Result<T>, tokio::time::error::Elapsed>,
+    limit: Duration,
+) -> trc::Result<T> {
+    match result {
+        Ok(result) => result,
+        Err(_) => {
+            drop(Object::take(conn));
+            Err(query_timeout_error(limit))
+        }
+    }
+}
+
 #[inline(always)]
 pub(crate) fn into_pool_error(err: deadpool_postgres::PoolError) -> trc::Error {
     match err {

crates/store/src/backend/postgres/read.rs

@@ -2,9 +2,11 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use super::{PostgresStore, into_error, is_timeout_error};
+use super::{PostgresStore, bounded, into_error, is_timeout_error, query_timeout_error};
 use crate::{
     Deserialize, IterateParams, Key, ValueKey, backend::postgres::into_pool_error,
     write::ValueClass,
@@ -17,40 +19,50 @@ impl PostgresStore {
         U: Deserialize + 'static,
     {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prepare_cached(&format!(
+        let result = tokio::time::timeout(limit, async {
-                "SELECT v FROM {} WHERE k = $1",
+            let s = conn
-                char::from(key.subspace())
+                .prepare_cached(&format!(
-            ))
+                    "SELECT v FROM {} WHERE k = $1",
-            .await
+                    char::from(key.subspace())
-            .map_err(into_error)?;
+                ))
-        let key = key.serialize(0);
+                .await
-        conn.query_opt(&s, &[&key])
+                .map_err(into_error)?;
-            .await
+            let key = key.serialize(0);
-            .map_err(into_error)
+            conn.query_opt(&s, &[&key])
-            .and_then(|r| {
+                .await
-                if let Some(r) = r {
+                .map_err(into_error)
-                    Ok(Some(U::deserialize_with_key(&key, r.get(0))?))
+                .and_then(|r| {
-                } else {
+                    if let Some(r) = r {
-                    Ok(None)
+                        Ok(Some(U::deserialize_with_key(&key, r.get(0))?))
-                }
+                    } else {
-            })
+                        Ok(None)
+                    }
+                })
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn key_exists(&self, key: impl Key) -> trc::Result<bool> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prepare_cached(&format!(
+        let result = tokio::time::timeout(limit, async {
-                "SELECT 1 FROM {} WHERE k = $1",
+            let s = conn
-                char::from(key.subspace())
+                .prepare_cached(&format!(
-            ))
+                    "SELECT 1 FROM {} WHERE k = $1",
-            .await
+                    char::from(key.subspace())
-            .map_err(into_error)?;
+                ))
-        let key = key.serialize(0);
+                .await
-        conn.query_opt(&s, &[&key])
+                .map_err(into_error)?;
-            .await
+            let key = key.serialize(0);
-            .map_err(into_error)
+            conn.query_opt(&s, &[&key])
-            .map(|r| r.is_some())
+                .await
+                .map_err(into_error)
+                .map(|r| r.is_some())
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub(crate) async fn iterate<T: Key>(
@@ -64,44 +76,65 @@ impl PostgresStore {
         let end = params.end.serialize(0);
         let keys = if params.values { "k, v" } else { "k" };
 
-        let s = conn
+        // inbuxa: a scan may run for hours, so the query limit bounds each
-            .prepare_cached(&match (params.first, params.ascending) {
+        // wait for the database (preparing, the query starting, the next
-                (true, true) => {
+        // row) rather than the scan. A wait that runs out closes the
-                    format!(
+        // connection.
-                        "SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k ASC LIMIT 1"
+        let limit = self.timeouts.query;
-                    )
+        let query = match (params.first, params.ascending) {
-                }
+            (true, true) => {
-                (true, false) => {
+                format!(
-                    format!(
+                    "SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k ASC LIMIT 1"
+                )
+            }
+            (true, false) => {
+                format!(
                     "SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k DESC LIMIT 1"
                 )
-                }
+            }
-                (false, true) => {
+            (false, true) => {
-                    format!("SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k ASC")
+                format!("SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k ASC")
-                }
+            }
-                (false, false) => {
+            (false, false) => {
-                    format!("SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k DESC")
+                format!("SELECT {keys} FROM {table} WHERE k >= $1 AND k <= $2 ORDER BY k DESC")
-                }
+            }
-            })
+        };
-            .await.map_err(into_error)?;
+        let s = match tokio::time::timeout(limit, conn.prepare_cached(&query)).await {
+            Ok(s) => s.map_err(into_error)?,
+            Err(_) => {
+                drop(deadpool_postgres::Object::take(conn));
+                return Err(query_timeout_error(limit));
+            }
+        };
         let mut from = begin;
         let mut to = end;
         let mut resume_key: Option<Vec<u8>> = None;
 
+        let mut stalled = false;
+
         loop {
             let mut last_key = None;
             let mut timed_out = false;
 
             {
-                let rows = conn
+                let rows =
-                    .query_raw(&s, &[&from, &to])
+                    match tokio::time::timeout(limit, conn.query_raw(&s, &[&from, &to])).await {
-                    .await
+                        Ok(rows) => rows.map_err(into_error)?,
-                    .map_err(into_error)?;
+                        // Leaves the scan loop for the timeout below
+                        Err(_) => break,
+                    };
 
                 pin_mut!(rows);
 
                 loop {
-                    match rows.try_next().await {
+                    let next = match tokio::time::timeout(limit, rows.try_next()).await {
+                        Ok(next) => next,
+                        Err(_) => {
+                            stalled = true;
+                            break;
+                        }
+                    };
+                    match next {
                         Ok(Some(row)) => {
                             let key = row.try_get::<_, &[u8]>(0).map_err(into_error)?;
                             let value = if params.values {
@@ -132,6 +165,10 @@ impl PostgresStore {
                 }
             }
 
+            if stalled {
+                break;
+            }
+
             match last_key {
                 Some(last_key) if timed_out => {
                     if params.ascending {
@@ -144,6 +181,9 @@ impl PostgresStore {
                 _ => return Ok(()),
             }
         }
+
+        drop(deadpool_postgres::Object::take(conn));
+        Err(query_timeout_error(limit))
     }
 
     pub(crate) async fn get_counter(
@@ -155,14 +195,19 @@ impl PostgresStore {
         let key = key.serialize(0);
 
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.query;
-            .prepare_cached(&format!("SELECT v FROM {table} WHERE k = $1"))
+        let result = tokio::time::timeout(limit, async {
-            .await
+            let s = conn
-            .map_err(into_error)?;
+                .prepare_cached(&format!("SELECT v FROM {table} WHERE k = $1"))
-        match conn.query_opt(&s, &[&key]).await {
+                .await
-            Ok(Some(row)) => row.try_get(0).map_err(into_error),
+                .map_err(into_error)?;
-            Ok(None) => Ok(0),
+            match conn.query_opt(&s, &[&key]).await {
-            Err(e) => Err(into_error(e)),
+                Ok(Some(row)) => row.try_get(0).map_err(into_error),
-        }
+                Ok(None) => Ok(0),
+                Err(e) => Err(into_error(e)),
+            }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }

crates/store/src/backend/postgres/search.rs

@@ -10,8 +10,8 @@ use crate::{
     backend::{
         MAX_TOKEN_LENGTH,
         postgres::{
-            DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, PostgresStore, PsqlSearchField, into_error,
+            DELETE_CHUNK_SIZE, MIN_DELETE_CHUNK_SIZE, PostgresStore, PsqlSearchField, bounded,
-            into_pool_error, is_timeout_error,
+            into_error, into_pool_error, is_timeout_error,
         },
     },
     search::{
@@ -36,125 +36,130 @@ impl PostgresStore {
 
     pub async fn index(&self, documents: Vec<IndexDocument>) -> trc::Result<()> {
         let mut conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let trx = conn
+        let limit = self.timeouts.query;
-            .build_transaction()
+        let result = tokio::time::timeout(limit, async {
-            .isolation_level(IsolationLevel::ReadCommitted)
+            let trx = conn
-            .start()
+                .build_transaction()
-            .await
+                .isolation_level(IsolationLevel::ReadCommitted)
-            .map_err(into_error)?;
+                .start()
+                .await
+                .map_err(into_error)?;
 
-        for document in documents {
+            for document in documents {
-            let index = document.index;
+                let index = document.index;
-            let primary_keys = index.primary_keys();
+                let primary_keys = index.primary_keys();
-            let all_fields = index.all_fields();
+                let all_fields = index.all_fields();
-            let fields = document.fields;
+                let fields = document.fields;
-            // inbuxa: keyword text (addresses, contact fields, ...) is split into
+                // inbuxa: keyword text (addresses, contact fields, ...) is split into
-            // words before it reaches the text parser, see keyword_terms().
+                // words before it reaches the text parser, see keyword_terms().
-            let keywords = primary_keys
+                let keywords = primary_keys
-                .iter()
+                    .iter()
-                .chain(all_fields)
+                    .chain(all_fields)
-                .map(|field| match fields.get(field) {
+                    .map(|field| match fields.get(field) {
-                    Some(SearchValue::Text {
+                        Some(SearchValue::Text {
-                        value,
+                            value,
-                        language: Language::None,
+                            language: Language::None,
-                    }) if field.is_text() => Some(keyword_terms(value)),
+                        }) if field.is_text() => Some(keyword_terms(value)),
-                    _ => None,
+                        _ => None,
-                })
+                    })
-                .collect::<Vec<_>>();
+                    .collect::<Vec<_>>();
-            let mut values = Vec::with_capacity(fields.len() + 2);
+                let mut values = Vec::with_capacity(fields.len() + 2);
-            let mut query = format!("INSERT INTO {} (", index.psql_table());
+                let mut query = format!("INSERT INTO {} (", index.psql_table());
 
-            for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
+                for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
-                if i > 0 {
+                    if i > 0 {
-                    query.push(',');
+                        query.push(',');
-                }
+                    }
-                query.push_str(field.column());
+                    query.push_str(field.column());
 
-                if let Some(sort_column) = field.sort_column() {
+                    if let Some(sort_column) = field.sort_column() {
-                    query.push(',');
+                        query.push(',');
-                    query.push_str(sort_column);
+                        query.push_str(sort_column);
-                }
+                    }
-            }
-
-            query.push_str(") VALUES (");
-
-            for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
-                if i > 0 {
-                    query.push(',');
                 }
 
-                if let Some(value) = fields.get(field) {
+                query.push_str(") VALUES (");
-                    let value_ref = format!("${}", values.len() + 1);
-                    let (text_len, language) = if let SearchValue::Text { value, language } = value
-                    {
-                        (value.len(), self.ts_config(language))
-                    } else {
-                        (0, PG_UNSTEMMED_LANG)
-                    };
 
-                    if let Some(keywords) = &keywords[i] {
+                for (i, field) in primary_keys.iter().chain(all_fields).enumerate() {
-                        let _ = write!(&mut query, "to_tsvector('{language}',{value_ref})");
+                    if i > 0 {
-                        values.push(keywords as &(dyn ToSql + Sync));
+                        query.push(',');
-                        if field.sort_column().is_some() {
+                    }
-                            let value_ref = format!("${}", values.len() + 1);
+
-                            if text_len > 255 {
+                    if let Some(value) = fields.get(field) {
-                                let _ = write!(&mut query, ",left({value_ref},255)");
+                        let value_ref = format!("${}", values.len() + 1);
+                        let (text_len, language) =
+                            if let SearchValue::Text { value, language } = value {
+                                (value.len(), self.ts_config(language))
                             } else {
-                                let _ = write!(&mut query, ",{value_ref}");
+                                (0, PG_UNSTEMMED_LANG)
+                            };
+
+                        if let Some(keywords) = &keywords[i] {
+                            let _ = write!(&mut query, "to_tsvector('{language}',{value_ref})");
+                            values.push(keywords as &(dyn ToSql + Sync));
+                            if field.sort_column().is_some() {
+                                let value_ref = format!("${}", values.len() + 1);
+                                if text_len > 255 {
+                                    let _ = write!(&mut query, ",left({value_ref},255)");
+                                } else {
+                                    let _ = write!(&mut query, ",{value_ref}");
+                                }
+                                values.push(value as &(dyn ToSql + Sync));
                             }
-                            values.push(value as &(dyn ToSql + Sync));
+                            continue;
-                        }
+                        } else if field.is_text() {
-                        continue;
+                            let _ = write!(&mut query, "to_tsvector('{language}',{value_ref})");
-                    } else if field.is_text() {
+                        } else if text_len > 512 {
-                        let _ = write!(&mut query, "to_tsvector('{language}',{value_ref})");
+                            query.push_str("left(");
-                    } else if text_len > 512 {
-                        query.push_str("left(");
-                        query.push_str(&value_ref);
-                        query.push_str(",512)");
-                    } else {
-                        query.push_str(&value_ref);
-                    }
-
-                    if field.sort_column().is_some() {
-                        if text_len > 255 {
-                            query.push_str(",left(");
                             query.push_str(&value_ref);
-                            query.push_str(",255)");
+                            query.push_str(",512)");
                         } else {
-                            query.push(',');
                             query.push_str(&value_ref);
                         }
-                    }
 
-                    values.push(value as &(dyn ToSql + Sync));
+                        if field.sort_column().is_some() {
-                } else {
+                            if text_len > 255 {
-                    query.push_str("NULL");
+                                query.push_str(",left(");
-                    if field.sort_column().is_some() {
+                                query.push_str(&value_ref);
-                        query.push_str(",NULL");
+                                query.push_str(",255)");
+                            } else {
+                                query.push(',');
+                                query.push_str(&value_ref);
+                            }
+                        }
+
+                        values.push(value as &(dyn ToSql + Sync));
+                    } else {
+                        query.push_str("NULL");
+                        if field.sort_column().is_some() {
+                            query.push_str(",NULL");
+                        }
                     }
                 }
-            }
 
-            query.push_str(") ON CONFLICT (");
+                query.push_str(") ON CONFLICT (");
-            for (i, pkey) in primary_keys.iter().enumerate() {
+                for (i, pkey) in primary_keys.iter().enumerate() {
-                if i > 0 {
+                    if i > 0 {
-                    query.push(',');
+                        query.push(',');
+                    }
+                    query.push_str(pkey.column());
                 }
-                query.push_str(pkey.column());
+                query.push_str(") DO UPDATE SET ");
-            }
+                for (i, field) in all_fields.iter().enumerate() {
-            query.push_str(") DO UPDATE SET ");
+                    if i > 0 {
-            for (i, field) in all_fields.iter().enumerate() {
+                        query.push(',');
-                if i > 0 {
+                    }
-                    query.push(',');
+                    let column = field.column();
+                    let _ = write!(&mut query, "{column} = EXCLUDED.{column}");
                 }
-                let column = field.column();
+
-                let _ = write!(&mut query, "{column} = EXCLUDED.{column}");
+                trx.execute(&query, &values).await.map_err(into_error)?;
             }
 
-            trx.execute(&query, &values).await.map_err(into_error)?;
+            trx.commit().await.map_err(into_error)
-        }
+        })
-
+        .await;
-        trx.commit().await.map_err(into_error)
+        bounded(conn, result, limit)
     }
 
     pub async fn query<R: SearchDocumentId>(
@@ -170,16 +175,21 @@ impl PostgresStore {
             build_sort(&mut query, sort);
         }
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn.prepare_cached(&query).await.map_err(into_error)?;
+        let limit = self.timeouts.query;
+        let result = tokio::time::timeout(limit, async {
+            let s = conn.prepare_cached(&query).await.map_err(into_error)?;
 
-        conn.query(&s, params.as_slice())
+            conn.query(&s, params.as_slice())
-            .await
+                .await
-            .and_then(|rows| {
+                .and_then(|rows| {
-                rows.into_iter()
+                    rows.into_iter()
-                    .map(|row| row.try_get::<_, DocId>(0).map(|v| R::from_u64(v.0)))
+                        .map(|row| row.try_get::<_, DocId>(0).map(|v| R::from_u64(v.0)))
-                    .collect::<Result<Vec<R>, _>>()
+                        .collect::<Result<Vec<R>, _>>()
-            })
+                })
-            .map_err(into_error)
+                .map_err(into_error)
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     pub async fn unindex(&self, filter: SearchQuery) -> trc::Result<u64> {
@@ -189,40 +199,45 @@ impl PostgresStore {
         let params = self.build_filter(&mut where_clause, &filter.filters);
         let params = params.iter().map(SqlParam::as_sql).collect::<Vec<_>>();
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let s = conn
+        let limit = self.timeouts.maintenance;
-            .prepare_cached(&format!("DELETE FROM {table}{where_clause}"))
+        let result = tokio::time::timeout(limit, async {
-            .await
-            .map_err(into_error)?;
-
-        match conn.execute(&s, params.as_slice()).await {
-            Ok(deleted) => return Ok(deleted),
-            Err(err) if is_timeout_error(&err) => (),
-            Err(err) => return Err(into_error(err)),
-        }
-
-        let mut chunk_size = DELETE_CHUNK_SIZE;
-        let mut deleted = 0;
-
-        loop {
             let s = conn
-                .prepare_cached(&format!(
+                .prepare_cached(&format!("DELETE FROM {table}{where_clause}"))
-                    "DELETE FROM {table} WHERE ctid IN (SELECT ctid FROM {table}{where_clause} LIMIT {chunk_size})"
-                ))
                 .await
                 .map_err(into_error)?;
 
+            match conn.execute(&s, params.as_slice()).await {
+                Ok(deleted) => return Ok(deleted),
+                Err(err) if is_timeout_error(&err) => (),
+                Err(err) => return Err(into_error(err)),
+            }
+
+            let mut chunk_size = DELETE_CHUNK_SIZE;
+            let mut deleted = 0;
+
             loop {
-                match conn.execute(&s, params.as_slice()).await {
+                let s = conn
-                    Ok(0) => return Ok(deleted),
+                    .prepare_cached(&format!(
-                    Ok(affected) => deleted += affected,
+                        "DELETE FROM {table} WHERE ctid IN (SELECT ctid FROM {table}{where_clause} LIMIT {chunk_size})"
-                    Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
+                    ))
-                        chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                    .await
-                        break;
+                    .map_err(into_error)?;
+
+                loop {
+                    match conn.execute(&s, params.as_slice()).await {
+                        Ok(0) => return Ok(deleted),
+                        Ok(affected) => deleted += affected,
+                        Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
+                            chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                            break;
+                        }
+                        Err(err) => return Err(into_error(err)),
                     }
-                    Err(err) => return Err(into_error(err)),
                 }
             }
-        }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     fn build_filter<'x>(

crates/store/src/backend/postgres/write.rs

@@ -2,9 +2,11 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
-use super::{PostgresStore, into_error, is_timeout_error};
+use super::{PostgresStore, bounded, into_error, is_timeout_error};
 use crate::{
     IndexKey, Key, LogKey, SUBSPACE_COUNTER, SUBSPACE_IN_MEMORY_COUNTER, SUBSPACE_QUOTA,
     SUBSPACE_REGISTRY_IDX,
@@ -30,48 +32,53 @@ enum CommitError {
 impl PostgresStore {
     pub(crate) async fn write(&self, mut batch: Batch<'_>) -> trc::Result<AssignedIds> {
         let mut conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let start = Instant::now();
+        let limit = self.timeouts.query;
-        let mut retry_count = 0;
+        let result = tokio::time::timeout(limit, async {
+            let start = Instant::now();
+            let mut retry_count = 0;
 
-        loop {
+            loop {
-            match self.write_trx(&mut conn, &mut batch).await {
+                match self.write_trx(&mut conn, &mut batch).await {
-                Ok(result) => {
+                    Ok(result) => {
-                    return Ok(result);
+                        return Ok(result);
-                }
-                Err(err) => {
-                    match err {
-                        CommitError::Postgres(err) => match err.code() {
-                            Some(
-                                &SqlState::T_R_SERIALIZATION_FAILURE
-                                | &SqlState::T_R_DEADLOCK_DETECTED,
-                            ) if retry_count < MAX_COMMIT_ATTEMPTS
-                                && start.elapsed() < MAX_COMMIT_TIME => {}
-                            Some(&SqlState::UNIQUE_VIOLATION) => {
-                                return Err(trc::StoreEvent::AssertValueFailed
-                                    .into_err()
-                                    .reason("Unique violation")
-                                    .caused_by(trc::location!()));
-                            }
-                            _ => return Err(into_error(err)),
-                        },
-                        CommitError::Internal(err) => return Err(err),
-                        /*CommitError::Retry => {
-                            if retry_count > MAX_COMMIT_ATTEMPTS
-                                || start.elapsed() > MAX_COMMIT_TIME
-                            {
-                                return Err(trc::StoreEvent::AssertValueFailed
-                                    .into_err()
-                                    .caused_by(trc::location!()));
-                            }
-                        }*/
                     }
+                    Err(err) => {
+                        match err {
+                            CommitError::Postgres(err) => match err.code() {
+                                Some(
+                                    &SqlState::T_R_SERIALIZATION_FAILURE
+                                    | &SqlState::T_R_DEADLOCK_DETECTED,
+                                ) if retry_count < MAX_COMMIT_ATTEMPTS
+                                    && start.elapsed() < MAX_COMMIT_TIME => {}
+                                Some(&SqlState::UNIQUE_VIOLATION) => {
+                                    return Err(trc::StoreEvent::AssertValueFailed
+                                        .into_err()
+                                        .reason("Unique violation")
+                                        .caused_by(trc::location!()));
+                                }
+                                _ => return Err(into_error(err)),
+                            },
+                            CommitError::Internal(err) => return Err(err),
+                            /*CommitError::Retry => {
+                                if retry_count > MAX_COMMIT_ATTEMPTS
+                                    || start.elapsed() > MAX_COMMIT_TIME
+                                {
+                                    return Err(trc::StoreEvent::AssertValueFailed
+                                        .into_err()
+                                        .caused_by(trc::location!()));
+                                }
+                            }*/
+                        }
 
-                    let backoff = rand::rng().random_range(50..=300);
+                        let backoff = rand::rng().random_range(50..=300);
-                    tokio::time::sleep(Duration::from_millis(backoff)).await;
+                        tokio::time::sleep(Duration::from_millis(backoff)).await;
-                    retry_count += 1;
+                        retry_count += 1;
+                    }
                 }
             }
-        }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 
     async fn write_trx(
@@ -393,72 +400,81 @@ impl PostgresStore {
 
     pub(crate) async fn purge_store(&self) -> trc::Result<()> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
+        let limit = self.timeouts.maintenance;
+        let result = tokio::time::timeout(limit, async {
+            for subspace in [SUBSPACE_QUOTA, SUBSPACE_COUNTER, SUBSPACE_IN_MEMORY_COUNTER] {
+                purge_table(&conn, char::from(subspace)).await?;
+            }
 
-        for subspace in [SUBSPACE_QUOTA, SUBSPACE_COUNTER, SUBSPACE_IN_MEMORY_COUNTER] {
+            Ok(())
-            purge_table(&conn, char::from(subspace)).await?;
+        })
-        }
+        .await;
-
+        bounded(conn, result, limit)
-        Ok(())
     }
 
     pub(crate) async fn delete_range(&self, from: impl Key, to: impl Key) -> trc::Result<()> {
         let conn = self.conn_pool.get().await.map_err(into_pool_error)?;
-        let table = char::from(from.subspace());
+        let limit = self.timeouts.maintenance;
-        let mut from = from.serialize(0);
+        let result = tokio::time::timeout(limit, async {
-        let to = to.serialize(0);
+            let table = char::from(from.subspace());
+            let mut from = from.serialize(0);
+            let to = to.serialize(0);
 
-        let delete = conn
+            let delete = conn
-            .prepare_cached(&format!("DELETE FROM {table} WHERE k >= $1 AND k < $2"))
+                .prepare_cached(&format!("DELETE FROM {table} WHERE k >= $1 AND k < $2"))
-            .await
-            .map_err(into_error)?;
-
-        match conn.execute(&delete, &[&from, &to]).await {
-            Ok(_) => return Ok(()),
-            Err(err) if is_timeout_error(&err) => (),
-            Err(err) => return Err(into_error(err)),
-        }
-
-        let mut chunk_size = DELETE_CHUNK_SIZE;
-
-        loop {
-            let boundary = conn
-                .prepare_cached(&format!(
-                    "SELECT k FROM {table} WHERE k >= $1 AND k < $2 ORDER BY k ASC LIMIT 1 OFFSET {chunk_size}"
-                ))
                 .await
                 .map_err(into_error)?;
 
+            match conn.execute(&delete, &[&from, &to]).await {
+                Ok(_) => return Ok(()),
+                Err(err) if is_timeout_error(&err) => (),
+                Err(err) => return Err(into_error(err)),
+            }
+
+            let mut chunk_size = DELETE_CHUNK_SIZE;
+
             loop {
-                let next = match conn.query_opt(&boundary, &[&from, &to]).await {
+                let boundary = conn
-                    Ok(next) => match next {
+                    .prepare_cached(&format!(
-                        Some(row) => Some(row.try_get::<_, Vec<u8>>(0).map_err(into_error)?),
+                        "SELECT k FROM {table} WHERE k >= $1 AND k < $2 ORDER BY k ASC LIMIT 1 OFFSET {chunk_size}"
-                        None => None,
+                    ))
-                    },
-                    Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
-                        chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
-                        break;
-                    }
-                    Err(err) => return Err(into_error(err)),
-                };
-
-                match conn
-                    .execute(&delete, &[&from, next.as_ref().unwrap_or(&to)])
                     .await
-                {
+                    .map_err(into_error)?;
-                    Ok(_) => (),
-                    Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
-                        chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
-                        break;
-                    }
-                    Err(err) => return Err(into_error(err)),
-                }
 
-                match next {
+                loop {
-                    Some(next) => from = next,
+                    let next = match conn.query_opt(&boundary, &[&from, &to]).await {
-                    None => return Ok(()),
+                        Ok(next) => match next {
+                            Some(row) => Some(row.try_get::<_, Vec<u8>>(0).map_err(into_error)?),
+                            None => None,
+                        },
+                        Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
+                            chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                            break;
+                        }
+                        Err(err) => return Err(into_error(err)),
+                    };
+
+                    match conn
+                        .execute(&delete, &[&from, next.as_ref().unwrap_or(&to)])
+                        .await
+                    {
+                        Ok(_) => (),
+                        Err(err) if is_timeout_error(&err) && chunk_size > MIN_DELETE_CHUNK_SIZE => {
+                            chunk_size = (chunk_size / 2).max(MIN_DELETE_CHUNK_SIZE);
+                            break;
+                        }
+                        Err(err) => return Err(into_error(err)),
+                    }
+
+                    match next {
+                        Some(next) => from = next,
+                        None => return Ok(()),
+                    }
                 }
             }
-        }
+        })
+        .await;
+        bounded(conn, result, limit)
     }
 }
 

crates/store/src/backend/query_timeout.rs

@@ -0,0 +1,77 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! Client-side limits on SQL queries.
+//!
+//! The pool timeouts bound getting a connection, not using one. A database
+//! that stops answering while the TCP connection stays up (a paused
+//! container, a hung server whose kernel still acknowledges keepalives)
+//! left a query on a checked-out connection waiting for as long as it took.
+//! A server-side statement_timeout can't help there: the server that would
+//! enforce it is the one not answering. So each operation on a PostgreSQL
+//! or MySQL connection runs under a time limit here, and a connection whose
+//! operation ran out is closed rather than put back in the pool, since its
+//! protocol state is unknown.
+//!
+//! Two limits:
+//! - `query`, two minutes, for request-path work: reads, writes, blob
+//!   transfers, search queries and document indexing. Those take
+//!   milliseconds; two minutes leaves room for a large blob over a slow
+//!   link and still ends a hang.
+//! - `maintenance`, thirty minutes, for work that legitimately runs long in
+//!   one statement: range deletes (account removal, purges), unindexing,
+//!   and creating tables and indexes at startup.
+//!
+//! Iterating over a range (exports, reindexing, maintenance scans) can run
+//! for hours, so there the `query` limit applies to each wait for the next
+//! row instead of the whole scan.
+
+use std::time::Duration;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub struct QueryTimeouts {
+    pub query: Duration,
+    pub maintenance: Duration,
+}
+
+impl QueryTimeouts {
+    pub const QUERY: Duration = Duration::from_secs(120);
+    pub const MAINTENANCE: Duration = Duration::from_secs(30 * 60);
+}
+
+impl Default for QueryTimeouts {
+    fn default() -> Self {
+        Self {
+            query: Self::QUERY,
+            maintenance: Self::MAINTENANCE,
+        }
+    }
+}
+
+#[cfg(feature = "test_mode")]
+impl crate::Store {
+    /// Sets the query limits of a SQL store that was just built (tests only:
+    /// the limits aren't configurable).
+    pub fn with_query_timeouts(self, timeouts: QueryTimeouts) -> Self {
+        match self {
+            #[cfg(feature = "postgres")]
+            crate::Store::PostgreSQL(mut store) => {
+                std::sync::Arc::get_mut(&mut store)
+                    .expect("store already shared")
+                    .timeouts = timeouts;
+                crate::Store::PostgreSQL(store)
+            }
+            #[cfg(feature = "mysql")]
+            crate::Store::MySQL(mut store) => {
+                std::sync::Arc::get_mut(&mut store)
+                    .expect("store already shared")
+                    .timeouts = timeouts;
+                crate::Store::MySQL(store)
+            }
+            store => store,
+        }
+    }
+}

crates/store/src/backend/redis/lookup.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use super::{RedisPool, RedisStore, into_error};
@@ -79,6 +81,30 @@ impl RedisStore {
         }
     }
 
+    // inbuxa: see InMemoryStore::renew_lock
+    pub async fn renew_lock(&self, key: &[u8], expires: u64) -> trc::Result<bool> {
+        match &self.pool {
+            RedisPool::Single(pool) => {
+                with_conn(pool, async |conn| {
+                    Self::renew_lock_(conn, key, expires).await
+                })
+                .await
+            }
+            RedisPool::Cluster(pool) => {
+                with_conn(pool, async |conn| {
+                    Self::renew_lock_(conn, key, expires).await
+                })
+                .await
+            }
+            RedisPool::Sentinel(pool) => {
+                with_conn(pool, async |conn| {
+                    Self::renew_lock_(conn, key, expires).await
+                })
+                .await
+            }
+        }
+    }
+
     pub async fn key_delete(&self, key: &[u8]) -> trc::Result<()> {
         match &self.pool {
             RedisPool::Single(pool) => {
@@ -226,6 +252,22 @@ impl RedisStore {
             .map(|reply| reply.is_some())
     }
 
+    async fn renew_lock_(
+        conn: &mut impl AsyncCommands,
+        key: &[u8],
+        expires: u64,
+    ) -> RedisResult<bool> {
+        redis::cmd("SET")
+            .arg(key)
+            .arg(now() + expires)
+            .arg("XX")
+            .arg("EX")
+            .arg(expires as i64)
+            .query_async::<Option<String>>(conn)
+            .await
+            .map(|reply| reply.is_some())
+    }
+
     async fn key_delete_(conn: &mut impl AsyncCommands, key: &[u8]) -> RedisResult<()> {
         conn.del(key).await
     }

crates/store/src/dispatch/lookup.rs

@@ -401,6 +401,57 @@ impl InMemoryStore {
         }
     }
 
+    /// inbuxa: extends a lock this node holds to `duration` seconds from now.
+    /// Returns false when the lock is gone or has expired: it may have been
+    /// taken by someone else since, so it is left alone.
+    pub async fn renew_lock(&self, prefix: u8, key: &[u8], duration: u64) -> trc::Result<bool> {
+        match self {
+            InMemoryStore::Store(store) => {
+                let key = KeyValue::<()>::build_key(prefix, key);
+                let key = ValueClass::InMemory(InMemoryClass::Key(key));
+                let Some(lock_expiry) = store
+                    .get_value::<u64>(ValueKey::from(key.clone()))
+                    .await
+                    .caused_by(trc::location!())?
+                else {
+                    return Ok(false);
+                };
+                let now = now();
+                if lock_expiry <= now {
+                    return Ok(false);
+                }
+
+                let mut batch = BatchBuilder::new();
+                batch.assert_value(key.clone(), AssertValue::U64(lock_expiry));
+                batch.set(key, (now + duration).serialize());
+                match store.write(batch.build_all()).await {
+                    Ok(_) => Ok(true),
+                    Err(err) if err.is_assertion_failure() => Ok(false),
+                    Err(err) => Err(err
+                        .details("Failed to renew lock.")
+                        .caused_by(trc::location!())),
+                }
+            }
+            InMemoryStore::Sharded(store) => {
+                Box::pin(
+                    store
+                        .member(&KeyValue::<()>::build_key(prefix, key))
+                        .renew_lock(prefix, key, duration),
+                )
+                .await
+            }
+            #[cfg(feature = "redis")]
+            InMemoryStore::Redis(store) => {
+                store
+                    .renew_lock(&KeyValue::<()>::build_key(prefix, key), duration)
+                    .await
+            }
+            InMemoryStore::Static(_) | InMemoryStore::Http(_) => {
+                Err(trc::StoreEvent::NotSupported.into_err())
+            }
+        }
+    }
+
     pub async fn remove_lock(&self, prefix: u8, key: &[u8]) -> trc::Result<()> {
         self.key_delete(KeyValue::<()>::build_key(prefix, key))
             .await

tests/src/cluster/broadcast.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::{
@@ -11,6 +13,7 @@ use crate::{
         server::TestServerBuilder,
     },
 };
+use common::BuildServer;
 use imap_proto::ResponseType;
 use registry::{
     schema::{
@@ -18,7 +21,8 @@ use registry::{
         prelude::{ObjectType, Property, SocketAddr},
         structs::{
             ClusterListenerGroup, ClusterListenerGroupProperties, ClusterRole, ClusterTaskGroup,
-            Coordinator, Imap, NatsCoordinator, NetworkListener, RedisStore,
+            Coordinator, Imap, MtaDeliverySchedule, MtaVirtualQueue, NatsCoordinator,
+            NetworkListener, RedisStore,
         },
     },
     types::map::Map,
@@ -209,6 +213,45 @@ pub async fn cluster_tests() {
         Some("John Doe")
     );
 
+    // inbuxa: a settings write applies on every node, no ReloadSettings
+    let queue_id = admin
+        .registry_create_object(MtaVirtualQueue {
+            name: "clusterq".into(),
+            threads_per_node: 1,
+            description: None,
+        })
+        .await;
+    admin
+        .registry_create_object(MtaDeliverySchedule {
+            name: "cluster-autoreload".into(),
+            queue_id,
+            ..Default::default()
+        })
+        .await;
+    for (node_id, test) in servers.iter().enumerate() {
+        let started = std::time::Instant::now();
+        while !test
+            .server
+            .inner
+            .build_server()
+            .core
+            .smtp
+            .queue
+            .queue_strategy
+            .contains_key("cluster-autoreload")
+        {
+            assert!(
+                started.elapsed() < std::time::Duration::from_secs(5),
+                "node {node_id} didn't pick up the new delivery schedule"
+            );
+            tokio::time::sleep(std::time::Duration::from_millis(50)).await;
+        }
+        println!(
+            "Node {node_id} has the new delivery schedule after {} ms",
+            started.elapsed().as_millis()
+        );
+    }
+
     // Run IMAP idle tests across nodes
     let mut node1_client = imap_client("[email protected]", "this is john's secret", 1).await;
     let mut node2_client = imap_client("[email protected]", "this is john's secret", 2).await;

tests/src/cluster/live_roles.rs

@@ -0,0 +1,297 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! A node follows edits to its cluster role without a restart: outbound
+//! delivery and report tasks start when the role gains outboundMta and stop
+//! when it loses it. Upstream decided at boot whether the queue, report and
+//! task managers ran at all. Needs a store the seed and the node can share
+//! (STORE=PostgreSql or MySql).
+
+use crate::utils::server::{TestServer, TestServerBuilder};
+use common::Server;
+use registry::{
+    schema::{
+        enums::ClusterTaskType,
+        prelude::{Object, ObjectType},
+        structs::{
+            ClusterListenerGroup, ClusterRole, ClusterTaskGroup, ClusterTaskGroupProperties, Task,
+            TaskStatus, TaskTlsReport,
+        },
+    },
+    types::{id::ObjectId, map::Map},
+};
+use smtp::{
+    queue::{Message, Status},
+    reporting::send::MtaReportSend,
+};
+use std::time::{Duration, Instant};
+use store::{
+    Deserialize, IterateParams, ValueKey,
+    registry::write::{RegistryWrite, RegistryWriteResult},
+    write::{AlignedBytes, Archive, BatchBuilder, QueueClass, TaskQueueClass, ValueClass},
+};
+use types::id::Id;
+use utils::snowflake::SnowflakeIdGenerator;
+
+const BUSY_ROLE: &str = "live_role_busy";
+const IDLE_ROLE: &str = "live_role_idle";
+const WITH: &[ClusterTaskType] = &[
+    ClusterTaskType::PushNotifications,
+    ClusterTaskType::OutboundMta,
+];
+const WITHOUT: &[ClusterTaskType] = &[ClusterTaskType::PushNotifications];
+const RCPT_DOMAIN: &str = "live-role.invalid";
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn live_role_tests() {
+    if matches!(
+        std::env::var("STORE").as_deref(),
+        Ok("RocksDb" | "Sqlite") | Err(_)
+    ) {
+        println!("Skipping live role tests: they need a store the nodes can share.");
+        return;
+    }
+    println!(
+        "Running live role tests on {}...",
+        std::env::var("STORE").unwrap_or_default()
+    );
+
+    // Two roles: one with outboundMta, one with no task type at all
+    let seed = TestServerBuilder::new("live_roles_seed").await;
+    let busy_id = seed.insert_object(role(BUSY_ROLE, WITH)).await;
+    let idle_id = seed.insert_object(role(IDLE_ROLE, WITHOUT)).await;
+    let seed = seed.disable_services().build().await;
+    let registry = seed.server.clone();
+
+    // 1. The rehearsal case: a node started with outboundMta has it taken
+    // away. Upstream kept delivering, report messages included, until a
+    // restart.
+    let node = start_node("live_roles_busy", BUSY_ROLE).await;
+    let server = node.server.clone();
+    assert!(server.core.network.roles.outbound_mta);
+    let (msg, task) = queue_work(&server, "busy-before").await;
+    assert_runs(&server, &msg, task).await;
+
+    set_role(&registry, &server, busy_id, role(BUSY_ROLE, WITHOUT)).await;
+    let (msg, task) = queue_work(&server, "busy-off").await;
+    assert_idle(&server, &msg, task).await;
+
+    // Given back, it takes up the work left waiting
+    set_role(&registry, &server, busy_id, role(BUSY_ROLE, WITH)).await;
+    assert_runs(&server, &msg, task).await;
+
+    // Off again, so it leaves the next node's work alone
+    set_role(&registry, &server, busy_id, role(BUSY_ROLE, WITHOUT)).await;
+
+    // 2. A node started with no task type at all gains outboundMta.
+    // Upstream never started its queue, report or task manager, so the
+    // role did nothing until a restart.
+    let node2 = start_node("live_roles_idle", IDLE_ROLE).await;
+    let server2 = node2.server.clone();
+    assert!(!server2.core.network.roles.outbound_mta);
+    let (msg, task) = queue_work(&server2, "idle-off").await;
+    assert_idle(&server2, &msg, task).await;
+    set_role(&registry, &server2, idle_id, role(IDLE_ROLE, WITH)).await;
+    assert_runs(&server2, &msg, task).await;
+    set_role(&registry, &server2, idle_id, role(IDLE_ROLE, WITHOUT)).await;
+
+    if seed.is_reset() {
+        seed.temp_dir.delete();
+        node.temp_dir.delete();
+        node2.temp_dir.delete();
+    }
+}
+
+async fn start_node(name: &str, role: &str) -> TestServer {
+    TestServerBuilder::new_with_role(
+        name,
+        format!("{name}.example.com").replace('_', "-"),
+        Some(role.into()),
+        false,
+    )
+    .await
+    .build_with_opts(false)
+    .await
+}
+
+/// Neither the message nor the report task is touched.
+async fn assert_idle(server: &Server, msg: &str, task: u64) {
+    tokio::time::sleep(Duration::from_secs(4)).await;
+    server.notify_task_queue();
+    tokio::time::sleep(Duration::from_secs(1)).await;
+    assert!(
+        !attempted(server, msg).await,
+        "delivery attempted without outboundMta"
+    );
+    assert!(
+        is_pending(server, task).await,
+        "report task claimed without outboundMta"
+    );
+}
+
+/// Delivery of the message is attempted and the report task runs.
+async fn assert_runs(server: &Server, msg: &str, task: u64) {
+    wait_for(Duration::from_secs(20), "message delivery attempt", || {
+        attempted(server, msg)
+    })
+    .await;
+    wait_for(Duration::from_secs(20), "report task to run", || async {
+        !is_pending(server, task).await
+    })
+    .await;
+}
+
+fn role(name: &str, tasks: &[ClusterTaskType]) -> ClusterRole {
+    ClusterRole {
+        name: name.into(),
+        description: None,
+        listeners: ClusterListenerGroup::EnableAll,
+        tasks: ClusterTaskGroup::EnableSome(ClusterTaskGroupProperties {
+            task_types: Map::new(tasks.to_vec()),
+        }),
+    }
+}
+
+/// Stores a new version of a role and reloads the node's settings, as a
+/// JMAP write to the role does.
+async fn set_role(registry: &Server, node: &Server, id: Id, new: ClusterRole) {
+    let enabled = matches!(&new.tasks, ClusterTaskGroup::EnableSome(group)
+        if group.task_types.iter().any(|t| *t == ClusterTaskType::OutboundMta));
+    let old = registry
+        .registry()
+        .get(ObjectId::new(ObjectType::ClusterRole, id))
+        .await
+        .unwrap()
+        .expect("role not found");
+    let new = Object::from(new);
+    let result = registry
+        .registry()
+        .write(RegistryWrite::update(id, &new, &old))
+        .await
+        .unwrap();
+    assert!(
+        matches!(result, RegistryWriteResult::Success(_)),
+        "role update refused"
+    );
+    assert_eq!(
+        node.reload_after_write(ObjectType::ClusterRole).await,
+        Some(Ok(()))
+    );
+    assert_eq!(
+        node.inner.shared_core.load().network.roles.outbound_mta,
+        enabled
+    );
+}
+
+/// Queues a message to an unreachable domain and schedules a TLS report
+/// task, both due now. Returns the recipient's local part and the task id.
+async fn queue_work(server: &Server, name: &str) -> (String, u64) {
+    let local = format!("{name}-{}", SnowflakeIdGenerator::global_id().unwrap());
+    let rcpt = format!("{local}@{RCPT_DOMAIN}");
+    server
+        .send_autogenerated(
+            "[email protected]",
+            [rcpt.as_str()].into_iter(),
+            format!(
+                "From: [email protected]\r\nTo: {rcpt}\r\n\
+                 Subject: live role test\r\n\r\nTest\r\n"
+            )
+            .into_bytes(),
+            None,
+            0,
+        )
+        .await;
+    assert!(
+        queued_recipient(server, &rcpt).await.is_some(),
+        "message to {rcpt} was not queued"
+    );
+
+    let task = SnowflakeIdGenerator::global_id().unwrap();
+    let mut batch = BatchBuilder::new();
+    batch.schedule_task_with_id(
+        task,
+        Task::TlsReport(TaskTlsReport {
+            report_id: u64::MAX.into(),
+            status: TaskStatus::now(),
+        }),
+    );
+    server.store().write(batch.build_all()).await.unwrap();
+    server.notify_task_queue();
+
+    (rcpt, task)
+}
+
+/// Whether delivery to `rcpt` was tried: the message is gone, or its
+/// recipient is no longer scheduled or has a retry count.
+async fn attempted(server: &Server, rcpt: &str) -> bool {
+    match queued_recipient(server, rcpt).await {
+        None => true,
+        Some((status_scheduled, retries)) => !status_scheduled || retries > 0,
+    }
+}
+
+/// The queued recipient `rcpt`: whether it is still scheduled, and how many
+/// times delivery was retried.
+async fn queued_recipient(server: &Server, rcpt: &str) -> Option<(bool, u32)> {
+    let mut found = None;
+    server
+        .store()
+        .iterate(
+            IterateParams::new(
+                ValueKey::from(ValueClass::Queue(QueueClass::Message(0))),
+                ValueKey::from(ValueClass::Queue(QueueClass::Message(u64::MAX))),
+            ),
+            |_, value| {
+                let message = <Archive<AlignedBytes> as Deserialize>::deserialize(value)?
+                    .deserialize::<Message>()?;
+                if let Some(recipient) = message
+                    .recipients
+                    .iter()
+                    .find(|recipient| recipient.address.as_ref() == rcpt)
+                {
+                    found = Some((
+                        matches!(recipient.status, Status::Scheduled),
+                        recipient.retry.inner,
+                    ));
+                    return Ok(false);
+                }
+                Ok(true)
+            },
+        )
+        .await
+        .unwrap();
+    found
+}
+
+async fn is_pending(server: &Server, id: u64) -> bool {
+    matches!(
+        server
+            .store()
+            .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(
+                TaskQueueClass::Task { id },
+            )))
+            .await
+            .unwrap()
+            .map(|task| task.status().clone()),
+        Some(TaskStatus::Pending(_))
+    )
+}
+
+async fn wait_for<F, Fut>(within: Duration, what: &str, mut check: F)
+where
+    F: FnMut() -> Fut,
+    Fut: Future<Output = bool>,
+{
+    let started = Instant::now();
+    while !check().await {
+        assert!(
+            started.elapsed() < within,
+            "still waiting for the {what} after {:?}",
+            started.elapsed()
+        );
+        tokio::time::sleep(Duration::from_millis(250)).await;
+    }
+}

tests/src/cluster/mod.rs

@@ -7,6 +7,8 @@
  */
 
 pub mod broadcast;
+pub mod live_roles; // inbuxa: role edits apply without a restart
 #[cfg(feature = "nats")]
 pub mod coordinator; // inbuxa: coordinator reconnects
 pub mod stress;
+pub mod task_roles; // inbuxa: task types follow cluster roles

tests/src/cluster/task_roles.rs

@@ -0,0 +1,218 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! Two task managers with different cluster roles over one shared store:
+//! each runs only the task types its role allows, and a task one node may
+//! not run is left for the node that may. Needs a store both nodes can open
+//! (STORE=PostgreSql or MySql).
+
+use crate::utils::server::TestServerBuilder;
+use common::Server;
+use registry::{
+    schema::{
+        enums::{ClusterTaskType, IndexDocumentType},
+        structs::{
+            ClusterListenerGroup, ClusterRole, ClusterTaskGroup, ClusterTaskGroupProperties, Task,
+            TaskDnsManagement, TaskIndexDocument, TaskStatus, TaskTlsReport,
+        },
+    },
+    types::map::Map,
+};
+use std::time::{Duration, Instant};
+use store::{
+    ValueKey,
+    write::{BatchBuilder, TaskQueueClass, ValueClass},
+};
+use utils::snowflake::SnowflakeIdGenerator;
+
+const QUEUE_ROLE: &str = "tasks_queue";
+const INDEX_MTA_ROLE: &str = "tasks_index_mta";
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn task_role_tests() {
+    if matches!(
+        std::env::var("STORE").as_deref(),
+        Ok("RocksDb" | "Sqlite") | Err(_)
+    ) {
+        println!("Skipping task role tests: they need a store both nodes can open.");
+        return;
+    }
+    println!(
+        "Running task role tests on {}...",
+        std::env::var("STORE").unwrap_or_default()
+    );
+
+    // The roles, stored by a node that runs no services of its own (a node
+    // looks its role up when it starts)
+    let seed = TestServerBuilder::new("task_roles_seed")
+        .await
+        .with_object(role(QUEUE_ROLE, &[ClusterTaskType::TaskQueueProcessing]))
+        .await
+        .with_object(role(
+            INDEX_MTA_ROLE,
+            &[
+                ClusterTaskType::SearchIndexing,
+                ClusterTaskType::OutboundMta,
+            ],
+        ))
+        .await
+        .disable_services()
+        .build()
+        .await;
+
+    // Node A runs queue tasks (taskQueueProcessing) only
+    let node_a = TestServerBuilder::new_with_role(
+        "task_roles_a",
+        "node-a.example.com".into(),
+        Some(QUEUE_ROLE.into()),
+        false,
+    )
+    .await
+    .build_with_opts(false)
+    .await;
+    let server_a = node_a.server.clone();
+    let roles = &server_a.core.network.roles;
+    assert!(roles.task_manager && !roles.search_indexing && !roles.outbound_mta);
+
+    // A DNS task (taskQueueProcessing), an unindex task (searchIndexing) and
+    // a TLS report (outboundMta), all due now
+    let [dns, unindex, report] = new_task_ids();
+    let mut batch = BatchBuilder::new();
+    batch
+        .schedule_task_with_id(
+            dns,
+            Task::DnsManagement(TaskDnsManagement {
+                status: TaskStatus::now(),
+                ..Default::default()
+            }),
+        )
+        .schedule_task_with_id(
+            unindex,
+            Task::UnindexDocument(TaskIndexDocument {
+                account_id: 0u32.into(),
+                document_id: u32::MAX.into(),
+                document_type: IndexDocumentType::File,
+                status: TaskStatus::now(),
+            }),
+        )
+        .schedule_task_with_id(
+            report,
+            Task::TlsReport(TaskTlsReport {
+                report_id: u64::MAX.into(),
+                status: TaskStatus::now(),
+            }),
+        );
+    server_a.store().write(batch.build_all()).await.unwrap();
+    server_a.notify_task_queue();
+
+    // Node A runs the DNS task and leaves the other two alone. Upstream ran
+    // the TLS report here too: report tasks ran on any node with a task
+    // manager.
+    wait_until_run(&server_a, &[dns], Duration::from_secs(20)).await;
+    tokio::time::sleep(Duration::from_secs(3)).await;
+    server_a.notify_task_queue();
+    tokio::time::sleep(Duration::from_secs(2)).await;
+    assert!(
+        is_pending(&server_a, unindex).await,
+        "unindex ran on node A"
+    );
+    assert!(
+        is_pending(&server_a, report).await,
+        "TLS report ran on node A"
+    );
+
+    // Node B (search indexing and outbound MTA) comes up and picks up what
+    // node A left
+    let node_b = TestServerBuilder::new_with_role(
+        "task_roles_b",
+        "node-b.example.com".into(),
+        Some(INDEX_MTA_ROLE.into()),
+        false,
+    )
+    .await
+    .build_with_opts(false)
+    .await;
+    let server_b = node_b.server.clone();
+    let roles = &server_b.core.network.roles;
+    assert!(!roles.task_manager && roles.search_indexing && roles.outbound_mta);
+    server_b.notify_task_queue();
+    wait_until_run(&server_b, &[unindex, report], Duration::from_secs(20)).await;
+
+    // A queue task scheduled now still runs, on node A: node B may not
+    // claim it
+    let [dns] = new_task_ids();
+    let mut batch = BatchBuilder::new();
+    batch.schedule_task_with_id(
+        dns,
+        Task::DnsManagement(TaskDnsManagement {
+            status: TaskStatus::now(),
+            ..Default::default()
+        }),
+    );
+    server_b.store().write(batch.build_all()).await.unwrap();
+    server_b.notify_task_queue();
+    tokio::time::sleep(Duration::from_secs(3)).await;
+    assert!(is_pending(&server_b, dns).await, "DNS task ran on node B");
+    server_a.notify_task_queue();
+    wait_until_run(&server_a, &[dns], Duration::from_secs(20)).await;
+
+    if seed.is_reset() {
+        seed.temp_dir.delete();
+        node_a.temp_dir.delete();
+        node_b.temp_dir.delete();
+    }
+}
+
+fn role(name: &str, tasks: &[ClusterTaskType]) -> ClusterRole {
+    ClusterRole {
+        name: name.into(),
+        description: None,
+        listeners: ClusterListenerGroup::EnableAll,
+        tasks: ClusterTaskGroup::EnableSome(ClusterTaskGroupProperties {
+            task_types: Map::new(tasks.to_vec()),
+        }),
+    }
+}
+
+fn new_task_ids<const N: usize>() -> [u64; N] {
+    std::array::from_fn(|_| SnowflakeIdGenerator::global_id().unwrap())
+}
+
+/// Still due and never run: present, and pending.
+async fn is_pending(server: &Server, id: u64) -> bool {
+    matches!(
+        server
+            .store()
+            .get_value::<Task>(ValueKey::from(ValueClass::TaskQueue(
+                TaskQueueClass::Task { id },
+            )))
+            .await
+            .unwrap()
+            .map(|task| task.status().clone()),
+        Some(TaskStatus::Pending(_))
+    )
+}
+
+async fn wait_until_run(server: &Server, ids: &[u64], within: Duration) {
+    let started = Instant::now();
+    loop {
+        let mut left = 0;
+        for id in ids {
+            if is_pending(server, *id).await {
+                left += 1;
+            }
+        }
+        if left == 0 {
+            return;
+        }
+        assert!(
+            started.elapsed() < within,
+            "{left} task(s) still pending after {:?}",
+            started.elapsed()
+        );
+        tokio::time::sleep(Duration::from_millis(250)).await;
+    }
+}

tests/src/smtp/inbound/mod.rs

@@ -2,6 +2,8 @@
  * SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
  *
  * SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
+ *
+ * Modified by Coffey Labs in 2026 for INBUXA.
  */
 
 use crate::utils::server::TestServer;
@@ -40,15 +42,23 @@ pub mod vrfy;
 const EVENT_TIMEOUT: Duration = Duration::from_secs(5);
 
 impl TestServer {
+    // inbuxa: registry writes reload the settings, and each reload sends the
+    // queue a ReloadSettings; read_event, try_read_event and assert_no_events
+    // pass over those (expect_reload_settings still waits for one)
     pub async fn read_event(&mut self) -> QueueEvent {
-        if let Some(event) = self.queue_events.pop_front() {
+        while let Some(event) = self.queue_events.pop_front() {
-            return event;
+            if !event.is_reload_settings() {
+                return event;
+            }
         }
 
-        match tokio::time::timeout(EVENT_TIMEOUT, self.queue_rx.recv()).await {
+        loop {
-            Ok(Some(event)) => event,
+            match tokio::time::timeout(EVENT_TIMEOUT, self.queue_rx.recv()).await {
-            Ok(None) => panic!("Channel closed."),
+                Ok(Some(event)) if event.is_reload_settings() => (),
-            Err(_) => panic!("No queue event received."),
+                Ok(Some(event)) => return event,
+                Ok(None) => panic!("Channel closed."),
+                Err(_) => panic!("No queue event received."),
+            }
         }
     }
 
@@ -78,26 +88,39 @@ impl TestServer {
     }
 
     pub async fn try_read_event(&mut self) -> Option<QueueEvent> {
-        if let Some(event) = self.queue_events.pop_front() {
+        while let Some(event) = self.queue_events.pop_front() {
-            return Some(event);
+            if !event.is_reload_settings() {
+                return Some(event);
+            }
         }
 
-        match tokio::time::timeout(EVENT_TIMEOUT, self.queue_rx.recv()).await {
+        loop {
-            Ok(Some(event)) => Some(event),
+            match tokio::time::timeout(EVENT_TIMEOUT, self.queue_rx.recv()).await {
-            Ok(None) => panic!("Channel closed."),
+                Ok(Some(event)) if event.is_reload_settings() => (),
-            Err(_) => None,
+                Ok(Some(event)) => return Some(event),
+                Ok(None) => panic!("Channel closed."),
+                Err(_) => return None,
+            }
         }
     }
 
     pub fn assert_no_events(&mut self) {
-        if let Some(event) = self.queue_events.pop_front() {
+        if let Some(event) = self
+            .queue_events
+            .iter()
+            .find(|event| !event.is_reload_settings())
+        {
             panic!("Expected empty queue but got {event:?}");
         }
+        self.queue_events.clear();
 
-        match self.queue_rx.try_recv() {
+        loop {
-            Err(TryRecvError::Empty) => (),
+            match self.queue_rx.try_recv() {
-            Ok(event) => panic!("Expected empty queue but got {event:?}"),
+                Ok(event) if event.is_reload_settings() => (),
-            Err(err) => panic!("Queue error: {err:?}"),
+                Err(TryRecvError::Empty) => break,
+                Ok(event) => panic!("Expected empty queue but got {event:?}"),
+                Err(err) => panic!("Queue error: {err:?}"),
+            }
         }
     }
 

tests/src/store/mod.rs

@@ -10,6 +10,8 @@ pub mod blob;
 pub mod import_export;
 pub mod lookup;
 pub mod ops;
+#[cfg(any(feature = "postgres", feature = "mysql"))]
+pub mod pool_timeout; // inbuxa: SQL pools give up instead of hanging
 pub mod query;
 pub mod registry;
 #[cfg(feature = "postgres")]
@@ -19,6 +21,8 @@ pub mod replica_mysql; // inbuxa: read replicas on MySQL
 #[cfg(all(feature = "postgres", feature = "redis"))]
 pub mod replica_cluster; // inbuxa: read replicas across nodes
 pub mod scaleout; // inbuxa: scale-out storage
+#[cfg(feature = "postgres")]
+pub mod search_gin; // inbuxa: GIN indexes without a pending list
 #[cfg(any(feature = "postgres", feature = "mysql"))]
 pub mod sql_timeout;
 pub mod task_locks; // inbuxa: task locks across nodes

tests/src/store/pool_timeout.rs

@@ -0,0 +1,376 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! A database that accepts connections and then says nothing (a hung or
+//! half-dead server, a black-holed failover) gives a worker an error within
+//! the pool's timeouts. Upstream's pools had none, so the worker waited for
+//! good. No database is needed: a local listener that never answers plays
+//! the server.
+//!
+//! inbuxa: the same for a database that stops answering while connections
+//! are already open (a paused container): a query on a checked-out
+//! connection ends within the query limit, the store works again once the
+//! database is back, and /healthz/ready says 503 in between while
+//! /healthz/live stays 200. These need the local test databases; a proxy
+//! that can stop forwarding plays the pause.
+
+use registry::schema::structs::DataStore;
+use std::{
+    sync::{
+        Arc,
+        atomic::{AtomicBool, Ordering},
+    },
+    time::{Duration, Instant},
+};
+use store::{
+    IterateParams, Store, ValueKey,
+    backend::query_timeout::QueryTimeouts,
+    write::{BatchBuilder, ValueClass},
+};
+use tokio::{
+    io::{AsyncReadExt, AsyncWriteExt},
+    net::{TcpListener, TcpStream},
+};
+
+/// Accepts connections on a local port and never sends a byte.
+async fn silent_server() -> u16 {
+    let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
+    let port = listener.local_addr().unwrap().port();
+    tokio::spawn(async move {
+        let mut held = Vec::new();
+        while let Ok((socket, _)) = listener.accept().await {
+            held.push(socket);
+        }
+    });
+    port
+}
+
+/// Builds the store and reads a key; both must end, with an error for the
+/// read, well within `limit`.
+async fn assert_times_out(data_store: DataStore, limit: Duration) {
+    let started = Instant::now();
+    let result = tokio::time::timeout(limit, async {
+        match Store::build(data_store).await {
+            Ok(store) => store
+                .get_value::<u64>(ValueKey::from(ValueClass::Property(0)))
+                .await
+                .map(|_| ())
+                .map_err(|err| err.to_string()),
+            Err(err) => Err(err.to_string()),
+        }
+    })
+    .await;
+    let elapsed = started.elapsed();
+    match result {
+        Ok(Err(err)) => println!("Got {err} after {elapsed:?}"),
+        Ok(Ok(())) => panic!("a silent server answered?"),
+        Err(_) => panic!("still waiting for a connection after {elapsed:?}"),
+    }
+}
+
+#[cfg(feature = "postgres")]
+#[tokio::test(flavor = "multi_thread")]
+pub async fn postgres_pool_timeout() {
+    use registry::schema::structs::PostgreSqlStore;
+
+    let port = silent_server().await;
+    println!("Running PostgreSQL pool timeout test...");
+    // The store's own timeout bounds opening a connection, handshake
+    // included (tokio-postgres's connect_timeout covers only the TCP connect)
+    assert_times_out(
+        DataStore::PostgreSql(PostgreSqlStore {
+            host: "127.0.0.1".into(),
+            port: port as u64,
+            database: "none".into(),
+            timeout: Some(Duration::from_secs(2).into()),
+            use_tls: false,
+            ..Default::default()
+        }),
+        Duration::from_secs(20),
+    )
+    .await;
+}
+
+#[cfg(feature = "mysql")]
+#[tokio::test(flavor = "multi_thread")]
+pub async fn mysql_pool_timeout() {
+    use registry::schema::structs::MySqlStore;
+
+    let port = silent_server().await;
+    println!("Running MySQL pool timeout test...");
+    // mysql_async has no pool timeout; the store waits 30 s for a connection
+    assert_times_out(
+        DataStore::MySql(MySqlStore {
+            host: "127.0.0.1".into(),
+            port: port as u64,
+            database: "none".into(),
+            use_tls: false,
+            ..Default::default()
+        }),
+        Duration::from_secs(60),
+    )
+    .await;
+}
+
+/// A TCP proxy to a local port that can stop forwarding, in both
+/// directions, while keeping every connection open: a paused server whose
+/// kernel still keeps the connections up.
+struct PausableProxy {
+    port: u16,
+    paused: Arc<AtomicBool>,
+}
+
+impl PausableProxy {
+    async fn start(upstream: u16) -> Self {
+        let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
+        let port = listener.local_addr().unwrap().port();
+        let paused = Arc::new(AtomicBool::new(false));
+        let paused_ = paused.clone();
+        tokio::spawn(async move {
+            while let Ok((client, _)) = listener.accept().await {
+                let Ok(server) = TcpStream::connect(("127.0.0.1", upstream)).await else {
+                    continue;
+                };
+                let (client_rx, client_tx) = client.into_split();
+                let (server_rx, server_tx) = server.into_split();
+                tokio::spawn(forward(client_rx, server_tx, paused_.clone()));
+                tokio::spawn(forward(server_rx, client_tx, paused_.clone()));
+            }
+        });
+        PausableProxy { port, paused }
+    }
+
+    fn pause(&self, paused: bool) {
+        self.paused.store(paused, Ordering::SeqCst);
+    }
+}
+
+async fn forward(
+    mut from: tokio::net::tcp::OwnedReadHalf,
+    mut to: tokio::net::tcp::OwnedWriteHalf,
+    paused: Arc<AtomicBool>,
+) {
+    let mut buf = vec![0u8; 16384];
+    loop {
+        while paused.load(Ordering::SeqCst) {
+            tokio::time::sleep(Duration::from_millis(20)).await;
+        }
+        let n = match from.read(&mut buf).await {
+            Ok(0) | Err(_) => return,
+            Ok(n) => n,
+        };
+        // Hold what arrived while paused until the pause ends
+        while paused.load(Ordering::SeqCst) {
+            tokio::time::sleep(Duration::from_millis(20)).await;
+        }
+        if to.write_all(&buf[..n]).await.is_err() {
+            return;
+        }
+    }
+}
+
+const TEST_LIMITS: QueryTimeouts = QueryTimeouts {
+    query: Duration::from_secs(2),
+    maintenance: Duration::from_secs(3),
+};
+
+/// Opens `connections` pooled connections at once, so the operations that
+/// follow find one idle and check it out.
+async fn warm(store: &Store, connections: usize) {
+    let reads = (0..connections).map(|_| async {
+        store
+            .get_value::<u64>(ValueKey::from(ValueClass::Property(0)))
+            .await
+            .unwrap();
+    });
+    futures::future::join_all(reads).await;
+}
+
+/// With the database paused, reads, scans and writes on connections the
+/// pool already holds end in an error within the query limit; once it is
+/// back, the store works again.
+async fn assert_queries_time_out(store: Store, proxy: &PausableProxy) {
+    store.create_tables().await.unwrap();
+    warm(&store, 4).await;
+    // mysql_async resets a connection on its way back to the pool; let
+    // those finish, or the connections are stuck in the reset when the
+    // pause starts and the pool's own wait timeout answers instead
+    tokio::time::sleep(Duration::from_secs(1)).await;
+    proxy.pause(true);
+
+    let key = || ValueKey::from(ValueClass::Property(0));
+    let limit = TEST_LIMITS.query;
+    for (what, op) in [("read", 0), ("scan", 1), ("write", 2)] {
+        let started = Instant::now();
+        let result = tokio::time::timeout(Duration::from_secs(20), async {
+            match op {
+                0 => store.get_value::<u64>(key()).await.map(|_| ()),
+                1 => {
+                    store
+                        .iterate(
+                            IterateParams::new(
+                                ValueKey::from(ValueClass::Property(0)),
+                                ValueKey::from(ValueClass::Property(u8::MAX)),
+                            ),
+                            |_, _| Ok(true),
+                        )
+                        .await
+                }
+                _ => {
+                    let mut batch = BatchBuilder::new();
+                    batch
+                        .with_account_id(u32::MAX - 7)
+                        .with_collection(types::collection::Collection::Email)
+                        .with_document(0)
+                        .set(ValueClass::Property(0), 1u64.to_be_bytes().to_vec());
+                    store.write(batch.build_all()).await.map(|_| ())
+                }
+            }
+        })
+        .await;
+        let elapsed = started.elapsed();
+        match result {
+            Ok(Err(err)) => {
+                let err = format!("{err:?}");
+                println!("Paused database, {what}: {err} after {elapsed:?}");
+                assert!(err.contains("Query timed out"), "{what}: {err}");
+                assert!(
+                    elapsed >= limit && elapsed < limit * 3,
+                    "{what} ended after {elapsed:?}"
+                );
+            }
+            Ok(Ok(())) => panic!("{what} succeeded against a paused database"),
+            Err(_) => panic!("{what} still waiting after {elapsed:?}"),
+        }
+    }
+
+    proxy.pause(false);
+    tokio::time::timeout(Duration::from_secs(20), store.get_value::<u64>(key()))
+        .await
+        .expect("still waiting after the database came back")
+        .expect("the store didn't recover");
+}
+
+#[cfg(feature = "postgres")]
+#[tokio::test(flavor = "multi_thread")]
+pub async fn postgres_query_timeout() {
+    println!("Running PostgreSQL query timeout test...");
+    let DataStore::PostgreSql(mut config) =
+        crate::utils::storage::build_data_store("PostgreSql", "").await
+    else {
+        unreachable!()
+    };
+    let proxy = PausableProxy::start(config.port as u16).await;
+    config.host = "127.0.0.1".into();
+    config.port = proxy.port as u64;
+    // New connections through the paused proxy give up as quickly
+    config.timeout = Some(TEST_LIMITS.query.into());
+    let store = Store::build(DataStore::PostgreSql(config))
+        .await
+        .unwrap()
+        .with_query_timeouts(TEST_LIMITS);
+    assert_queries_time_out(store, &proxy).await;
+}
+
+#[cfg(feature = "mysql")]
+#[tokio::test(flavor = "multi_thread")]
+pub async fn mysql_query_timeout() {
+    println!("Running MySQL query timeout test...");
+    let DataStore::MySql(mut config) = crate::utils::storage::build_data_store("MySql", "").await
+    else {
+        unreachable!()
+    };
+    let proxy = PausableProxy::start(config.port as u16).await;
+    config.host = "127.0.0.1".into();
+    config.port = proxy.port as u64;
+    let store = Store::build(DataStore::MySql(config))
+        .await
+        .unwrap()
+        .with_query_timeouts(TEST_LIMITS);
+    assert_queries_time_out(store, &proxy).await;
+}
+
+/// /healthz/ready follows the data store; /healthz/live doesn't.
+#[cfg(feature = "postgres")]
+#[tokio::test(flavor = "multi_thread")]
+pub async fn postgres_readiness() {
+    use crate::utils::server::TestServerBuilder;
+    use registry::schema::enums::NetworkListenerProtocol;
+
+    const HTTP_PORT: u16 = 11_320;
+    if std::env::var("STORE").as_deref() != Ok("PostgreSql") {
+        println!("Skipping the readiness test: it runs with STORE=PostgreSql.");
+        return;
+    }
+    println!("Running readiness test...");
+
+    let test = TestServerBuilder::new("postgres_readiness")
+        .await
+        .with_listener(NetworkListenerProtocol::Http, "http", HTTP_PORT, true)
+        .await
+        .build()
+        .await;
+
+    // Point the running node's data store at the database through the proxy
+    let DataStore::PostgreSql(mut config) =
+        crate::utils::storage::build_data_store("PostgreSql", "").await
+    else {
+        unreachable!()
+    };
+    let proxy = PausableProxy::start(config.port as u16).await;
+    config.host = "127.0.0.1".into();
+    config.port = proxy.port as u64;
+    config.timeout = Some(TEST_LIMITS.query.into());
+    let store = Store::build(DataStore::PostgreSql(config))
+        .await
+        .unwrap()
+        .with_query_timeouts(TEST_LIMITS);
+    let inner = &test.server.inner;
+    let mut core = inner.shared_core.load_full().as_ref().clone();
+    core.storage.data = store;
+    inner.shared_core.store(Arc::new(core));
+
+    let health = |path: &'static str| async move {
+        reqwest::Client::builder()
+            .danger_accept_invalid_certs(true)
+            .timeout(Duration::from_secs(10))
+            .build()
+            .unwrap()
+            .get(format!("https://127.0.0.1:{HTTP_PORT}/healthz/{path}"))
+            .send()
+            .await
+            .unwrap()
+            .status()
+            .as_u16()
+    };
+    let wait_for = |path: &'static str, status: u16| async move {
+        let started = Instant::now();
+        loop {
+            let got = health(path).await;
+            if got == status {
+                println!("/healthz/{path}: {got} after {:?}", started.elapsed());
+                return;
+            }
+            assert!(
+                started.elapsed() < Duration::from_secs(20),
+                "/healthz/{path} still {got}, expected {status}"
+            );
+            tokio::time::sleep(Duration::from_millis(250)).await;
+        }
+    };
+
+    wait_for("ready", 200).await;
+    proxy.pause(true);
+    wait_for("ready", 503).await;
+    assert_eq!(health("live").await, 200);
+    proxy.pause(false);
+    wait_for("ready", 200).await;
+    assert_eq!(health("live").await, 200);
+
+    if test.is_reset() {
+        test.temp_dir.delete();
+    }
+}

tests/src/store/search_gin.rs

@@ -0,0 +1,159 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+//! PostgreSQL full-text GIN indexes are built with fastupdate off, and an
+//! index made earlier with the default is switched over at startup. With
+//! fastupdate on, new entries wait in a pending list that every search scans
+//! in full until VACUUM merges it.
+
+use crate::utils::storage::build_data_store;
+use registry::schema::structs::DataStore;
+use store::{Rows, SearchStore, Store};
+
+const SCHEMA: &str = "gin_fastupdate_test";
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn postgres_gin_fastupdate() {
+    println!("Running PostgreSQL GIN fastupdate test...");
+
+    // Work in a schema of our own so the shared search tables are untouched
+    let admin = Store::build(build_data_store("PostgreSql", "").await)
+        .await
+        .expect("Failed to connect to PostgreSQL");
+    for query in [
+        format!("DROP SCHEMA IF EXISTS {SCHEMA} CASCADE"),
+        format!("CREATE SCHEMA {SCHEMA}"),
+    ] {
+        admin.sql_query::<usize>(&query, vec![]).await.unwrap();
+    }
+    let DataStore::PostgreSql(mut config) = build_data_store("PostgreSql", "").await else {
+        unreachable!()
+    };
+    config.options = Some(format!("-c search_path={SCHEMA}"));
+    let store = Store::build(DataStore::PostgreSql(config))
+        .await
+        .expect("Failed to connect to PostgreSQL");
+    let search = SearchStore::Store(store.clone());
+
+    // A fresh schema
+    search.create_indexes().await.unwrap();
+    let indexes = gin_indexes(&admin).await;
+    assert!(
+        indexes.len() >= 4,
+        "expected the search GIN indexes, found {indexes:?}"
+    );
+    for (name, options) in &indexes {
+        assert!(
+            options.contains("fastupdate=off"),
+            "fresh index {name} has options {options:?}"
+        );
+    }
+
+    // A schema from before the change: the same indexes, made with the
+    // default fastupdate=on, and a pending list with something in it
+    for (name, _) in &indexes {
+        admin
+            .sql_query::<usize>(
+                &format!("ALTER INDEX {SCHEMA}.{name} RESET (fastupdate)"),
+                vec![],
+            )
+            .await
+            .unwrap();
+    }
+    for (name, options) in gin_indexes(&admin).await {
+        assert!(
+            !options.contains("fastupdate"),
+            "index {name} still has options {options:?}"
+        );
+    }
+    admin
+        .sql_query::<usize>(
+            &format!(
+                "INSERT INTO {SCHEMA}.s_email (accid, docid, subj, body) \
+                 SELECT 1, n, to_tsvector('simple', 'pending subject ' || n), \
+                 to_tsvector('simple', 'pending body text ' || n) \
+                 FROM generate_series(1, 500) n"
+            ),
+            vec![],
+        )
+        .await
+        .unwrap();
+    assert!(
+        pending_tuples(&admin, "gin_s_email_body").await > 0,
+        "no pending list to merge"
+    );
+
+    // Startup on the existing schema switches every index over and merges
+    // what was pending
+    search.create_indexes().await.unwrap();
+    for (name, options) in gin_indexes(&admin).await {
+        assert!(
+            options.contains("fastupdate=off"),
+            "existing index {name} has options {options:?} after startup"
+        );
+    }
+    assert_eq!(pending_tuples(&admin, "gin_s_email_body").await, 0);
+
+    // And a second startup changes nothing
+    search.create_indexes().await.unwrap();
+    for (name, options) in gin_indexes(&admin).await {
+        assert!(options.contains("fastupdate=off"), "{name}: {options:?}");
+    }
+
+    admin
+        .sql_query::<usize>(&format!("DROP SCHEMA {SCHEMA} CASCADE"), vec![])
+        .await
+        .unwrap();
+}
+
+/// The GIN indexes in the test schema with their reloptions.
+async fn gin_indexes(admin: &Store) -> Vec<(String, String)> {
+    admin
+        .sql_query::<Rows>(
+            &format!(
+                "SELECT c.relname::text, COALESCE(array_to_string(c.reloptions, ','), '') \
+                 FROM pg_class c JOIN pg_namespace n ON n.oid = c.relnamespace \
+                 JOIN pg_am a ON a.oid = c.relam \
+                 WHERE n.nspname = '{SCHEMA}' AND c.relkind = 'i' AND a.amname = 'gin' \
+                 ORDER BY 1"
+            ),
+            vec![],
+        )
+        .await
+        .unwrap()
+        .rows
+        .into_iter()
+        .map(|row| {
+            let mut values = row.values.into_iter();
+            (
+                values.next().unwrap().to_str().into_owned(),
+                values.next().unwrap().to_str().into_owned(),
+            )
+        })
+        .collect()
+}
+
+/// Tuples waiting in a GIN index's pending list (pgstattuple is a contrib
+/// extension the test database has).
+async fn pending_tuples(admin: &Store, index: &str) -> i64 {
+    admin
+        .sql_query::<usize>("CREATE EXTENSION IF NOT EXISTS pgstattuple", vec![])
+        .await
+        .unwrap();
+    admin
+        .sql_query::<Rows>(
+            &format!("SELECT pending_tuples FROM pgstatginindex('{SCHEMA}.{index}'::regclass)"),
+            vec![],
+        )
+        .await
+        .unwrap()
+        .rows
+        .into_iter()
+        .next()
+        .and_then(|row| row.values.into_iter().next())
+        .map(|value| value.to_str().parse::<i64>().unwrap())
+        .unwrap()
+}

tests/src/store/task_locks.rs

@@ -79,7 +79,33 @@ pub async fn task_lock_tests() {
         "ran before the other node's locks expired: {elapsed:?}"
     );
 
-    // 3. A graceful stop releases the locks this node holds: another node
+    // 3. A task that runs longer than a lock lifetime keeps its claim: the
+    // task manager renews the lease while this node holds it, and the claim
+    // ends when the task does. (Before, a lock simply lasted an hour.)
+    let [id] = new_task_ids(1)[..] else {
+        unreachable!()
+    };
+    assert!(server.try_lock_task(id).await, "claim {id}");
+    tokio::time::sleep(Duration::from_secs(LOCK_EXPIRY + LOCK_EXPIRY / 2)).await;
+    assert!(
+        !foreign_lock(&server, id, LOCK_EXPIRY).await,
+        "lease lapsed while the task ran"
+    );
+    server.remove_index_lock(id).await;
+    assert!(
+        foreign_lock(&server, id, LOCK_EXPIRY).await,
+        "released when the task ended"
+    );
+    let _ = server
+        .in_memory_store()
+        .remove_lock(KV_LOCK_TASK, &id.to_be_bytes())
+        .await;
+    assert!(
+        common::ipc::TaskLocks::DEFAULT_EXPIRY <= 5 * 60,
+        "a dead node's tasks wait no more than a few minutes"
+    );
+
+    // 4. A graceful stop releases the locks this node holds: another node
     // can claim those tasks at once, and this one claims nothing more
     let ids = new_task_ids(3);
     for id in &ids {

tests/src/system/auto_reload.rs

@@ -0,0 +1,220 @@
+/*
+ * SPDX-FileCopyrightText: 2026 Coffey Labs
+ *
+ * SPDX-License-Identifier: AGPL-3.0-only
+ */
+
+// inbuxa: a registry write to an object the running settings are built from
+// applies without an x:Action ReloadSettings, and the set response says so.
+
+use crate::utils::{
+    jmap::JmapResponse,
+    server::{TestServer, TestServerBuilder},
+};
+use common::BuildServer;
+use registry::{
+    schema::{
+        enums::TracingLevel,
+        prelude::ObjectType,
+        structs::{
+            AllowedIp, CertificateManagement, DkimManagement, DnsManagement, Domain, Expression,
+            MtaDeliverySchedule, MtaStageAuth, MtaVirtualQueue, Tracer, TracerStdout,
+        },
+    },
+    types::ipmask::IpAddrOrMask,
+};
+use serde_json::Value;
+
+#[tokio::test(flavor = "multi_thread")]
+pub async fn settings_reload_tests() {
+    let mut test = TestServerBuilder::new("settings_reload_tests")
+        .await
+        .with_default_listeners()
+        .await
+        .with_object(MtaStageAuth {
+            require: Expression {
+                else_: "false".to_string(),
+                ..Default::default()
+            },
+            ..Default::default()
+        })
+        .await
+        .build()
+        .await;
+
+    let admin = test
+        .create_user_account(
+            "admin",
+            "[email protected]",
+            "these_pretzels_are_making_me_thirsty",
+            &[],
+            "Admin",
+        )
+        .await;
+    test.account("admin")
+        .assign_roles_to_account(admin.id(), &["user", "system"])
+        .await;
+    test.insert_account(admin);
+
+    test_write_applies(&test).await;
+
+    if test.is_reset() {
+        test.temp_dir.delete();
+    }
+}
+
+async fn test_write_applies(test: &TestServer) {
+    println!("Running settings reload after registry writes...");
+    let admin = test.account("[email protected]");
+
+    // A delivery schedule is in use as soon as it is saved
+    let response = admin
+        .registry_create([MtaVirtualQueue {
+            name: "autorld".into(),
+            threads_per_node: 2,
+            description: None,
+        }])
+        .await;
+    assert_applied(&response);
+    let queue_id = response.created_id(0);
+    assert!(!has_schedule(test, "autoreload-schedule"));
+    let response = admin
+        .registry_create([MtaDeliverySchedule {
+            name: "autoreload-schedule".into(),
+            queue_id,
+            ..Default::default()
+        }])
+        .await;
+    assert_applied(&response);
+    assert!(has_schedule(test, "autoreload-schedule"));
+
+    // Destroyed, it's gone at once too
+    let schedule_id = response.created_id(0);
+    let response = admin
+        .registry_destroy(ObjectType::MtaDeliverySchedule, [schedule_id])
+        .await;
+    assert_applied(&response);
+    assert!(!has_schedule(test, "autoreload-schedule"));
+
+    // Concurrent writes all end up in the running settings
+    let names = (0..8)
+        .map(|i| format!("autoreload-{i}"))
+        .collect::<Vec<_>>();
+    let mut writes = Vec::new();
+    for name in &names {
+        writes.push(admin.registry_create([MtaDeliverySchedule {
+            name: name.clone(),
+            queue_id,
+            ..Default::default()
+        }]));
+    }
+    let mut schedule_ids = Vec::new();
+    for response in futures::future::join_all(writes).await {
+        assert_applied(&response);
+        schedule_ids.push(response.created_id(0));
+    }
+    for name in &names {
+        assert!(has_schedule(test, name), "{name} missing");
+    }
+    // Several objects in one request: one reload
+    let response = admin
+        .registry_destroy(ObjectType::MtaDeliverySchedule, schedule_ids.iter())
+        .await;
+    assert_applied(&response);
+    for name in &names {
+        assert!(!has_schedule(test, name), "{name} still present");
+    }
+
+    // A write whose reload fails is stored, and the response says the
+    // settings weren't reloaded: only one console tracer is allowed.
+    let response = admin
+        .registry_create([
+            Tracer::Stdout(TracerStdout {
+                enable: true,
+                level: TracingLevel::Error,
+                ..Default::default()
+            }),
+            Tracer::Stdout(TracerStdout {
+                enable: true,
+                level: TracingLevel::Error,
+                ..Default::default()
+            }),
+        ])
+        .await;
+    let reload = settings_reload(&response).expect("x:settingsReload missing");
+    assert_eq!(reload["applied"], Value::Bool(false), "{response:?}");
+    let description = reload["description"].as_str().unwrap_or_default();
+    assert!(
+        description.starts_with("Saved, but the running settings were not reloaded. ")
+            && description.contains("Only one console tracer is allowed"),
+        "{description}"
+    );
+    let tracer_ids = [response.created_id(0), response.created_id(1)];
+    let response = admin
+        .registry_destroy(ObjectType::Tracer, tracer_ids.iter())
+        .await;
+    assert_applied(&response);
+
+    // An allowed IP is live as soon as it is saved, and gone once
+    // destroyed. It lives in the core's security settings, which the
+    // blocked-IP reload it used to get doesn't rebuild.
+    let ip: std::net::IpAddr = "198.51.100.7".parse().unwrap();
+    assert!(!is_allowed(test, ip));
+    let response = admin
+        .registry_create([AllowedIp {
+            address: IpAddrOrMask::from_ip(ip),
+            reason: Some("autoreload".into()),
+            ..Default::default()
+        }])
+        .await;
+    assert_applied(&response);
+    assert!(
+        is_allowed(test, ip),
+        "allowed IP not in the running settings"
+    );
+    let allowed_id = response.created_id(0);
+    let response = admin
+        .registry_destroy(ObjectType::AllowedIp, [allowed_id])
+        .await;
+    assert_applied(&response);
+    assert!(!is_allowed(test, ip), "destroyed allowed IP still live");
+
+    // Data that isn't part of the running settings doesn't reload them
+    let response = admin
+        .registry_create([Domain {
+            name: "autoreload.example.org".into(),
+            certificate_management: CertificateManagement::Manual,
+            dns_management: DnsManagement::Manual,
+            dkim_management: DkimManagement::Manual,
+            ..Default::default()
+        }])
+        .await;
+    assert!(settings_reload(&response).is_none(), "{response:?}");
+}
+
+fn settings_reload(response: &JmapResponse) -> Option<&Value> {
+    response.pointer("/methodResponses/0/1/x:settingsReload")
+}
+
+fn assert_applied(response: &JmapResponse) {
+    assert_eq!(
+        settings_reload(response),
+        Some(&serde_json::json!({"applied": true})),
+        "{response:?}"
+    );
+}
+
+fn has_schedule(test: &TestServer, name: &str) -> bool {
+    test.server
+        .inner
+        .build_server()
+        .core
+        .smtp
+        .queue
+        .queue_strategy
+        .contains_key(name)
+}
+
+fn is_allowed(test: &TestServer, ip: std::net::IpAddr) -> bool {
+    test.server.inner.build_server().is_ip_allowed(ip)
+}

tests/src/system/mod.rs

@@ -11,6 +11,7 @@ pub mod authentication;
 pub mod ai;
 pub mod ai_calibration;
 pub mod authorization;
+pub mod auto_reload; // inbuxa: registry writes apply at once
 pub mod branding;
 pub mod crypto;
 pub mod delivery;