Reported on Discord (#inbuxa, faisal, 2026-11-20): outbound DMARC reports sit in the queue for hours with no earlier attempt and no 4xx, then go out on the first try, and the console's delivery-time graph shows multi-hour spikes (peaks around 2.5 h, an 18 min average against 2 s per attempt).
Cause: a deliberate hold, counted as delivery time.
send_report (crates/smtp/src/reporting/send.rs) queues each aggregate report (DMARC, dmarc.rs:490; TLS-RPT, tls.rs:242, both deliver_now = false) with its first attempt at a random time up to 3 hours later (random_range(0..10800), unchanged from upstream). That spreads reports out so they don't all reach receivers at once, and it's fine to keep. Failure reports (DMARC/SPF/DKIM) go out at once.
The delivery-time figure is Delivery Completed's Elapsed, now() - message.created (crates/smtp/src/outbound/delivery.rs), so the hold is counted as delivery time. The queue-size chart also counts these reports while they wait.
Mail delivery is fine; the dashboard reads a scheduled hold as slow delivery.
Wanted
Delivery time leaves out the planned hold: count scheduled reports from their first scheduled attempt rather than from when they were queued, or keep them out of the figure.
In the queue, a held report says it's a scheduled report and when it's due, so it doesn't look stuck (console queue view, if it doesn't already show the first attempt time).
No change to the hold itself.
Reported on Discord (#inbuxa, faisal, 2026-11-20): outbound DMARC reports sit in the queue for hours with no earlier attempt and no 4xx, then go out on the first try, and the console's delivery-time graph shows multi-hour spikes (peaks around 2.5 h, an 18 min average against 2 s per attempt).
**Cause: a deliberate hold, counted as delivery time.**
- `send_report` (`crates/smtp/src/reporting/send.rs`) queues each aggregate report (DMARC, `dmarc.rs:490`; TLS-RPT, `tls.rs:242`, both `deliver_now = false`) with its first attempt at a random time up to 3 hours later (`random_range(0..10800)`, unchanged from upstream). That spreads reports out so they don't all reach receivers at once, and it's fine to keep. Failure reports (DMARC/SPF/DKIM) go out at once.
- The delivery-time figure is `Delivery Completed`'s `Elapsed`, `now() - message.created` (`crates/smtp/src/outbound/delivery.rs`), so the hold is counted as delivery time. The queue-size chart also counts these reports while they wait.
Mail delivery is fine; the dashboard reads a scheduled hold as slow delivery.
**Wanted**
- Delivery time leaves out the planned hold: count scheduled reports from their first scheduled attempt rather than from when they were queued, or keep them out of the figure.
- In the queue, a held report says it's a scheduled report and when it's due, so it doesn't look stuck (console queue view, if it doesn't already show the first attempt time).
- No change to the hold itself.
jcoffey-dev
added this to the 27.1 milestone 2026-11-20 23:19:35 +00:00
Fixed in #518. Each aggregate report's planned hold is now recorded on the queued message, and delivery time counts from the end of the hold, so the delivery-time graph, the stored delivery metric and delivery history no longer include it. The hold itself is unchanged. The queue's Next Retry column already shows when a held report will go out.
Ships in 27.1-rc.38. Reports already queued when you update still count their hold until they go out, at most 3 hours.
Fixed in #518. Each aggregate report's planned hold is now recorded on the queued message, and delivery time counts from the end of the hold, so the delivery-time graph, the stored delivery metric and delivery history no longer include it. The hold itself is unchanged. The queue's Next Retry column already shows when a held report will go out.
Ships in 27.1-rc.38. Reports already queued when you update still count their hold until they go out, at most 3 hours.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Reported on Discord (#inbuxa, faisal, 2026-11-20): outbound DMARC reports sit in the queue for hours with no earlier attempt and no 4xx, then go out on the first try, and the console's delivery-time graph shows multi-hour spikes (peaks around 2.5 h, an 18 min average against 2 s per attempt).
Cause: a deliberate hold, counted as delivery time.
send_report(crates/smtp/src/reporting/send.rs) queues each aggregate report (DMARC,dmarc.rs:490; TLS-RPT,tls.rs:242, bothdeliver_now = false) with its first attempt at a random time up to 3 hours later (random_range(0..10800), unchanged from upstream). That spreads reports out so they don't all reach receivers at once, and it's fine to keep. Failure reports (DMARC/SPF/DKIM) go out at once.Delivery Completed'sElapsed,now() - message.created(crates/smtp/src/outbound/delivery.rs), so the hold is counted as delivery time. The queue-size chart also counts these reports while they wait.Mail delivery is fine; the dashboard reads a scheduled hold as slow delivery.
Wanted
Fixed in #518. Each aggregate report's planned hold is now recorded on the queued message, and delivery time counts from the end of the hold, so the delivery-time graph, the stored delivery metric and delivery history no longer include it. The hold itself is unchanged. The queue's Next Retry column already shows when a held report will go out.
Ships in 27.1-rc.38. Reports already queued when you update still count their hold until they go out, at most 3 hours.