Delivery Queue Processor Only Runs on Recovery, Causing Multi-Hour Delays #21722
Description
Summary
Messages enqueued by cron jobs sit in the delivery queue for hours because the delivery processor only runs during gateway restart/recovery, not continuously.
Environment
- OpenClaw version: 2026.2.19-2
- OS: Ubuntu Linux 6.8.0-90-generic
- Gateway: systemd service, loopback mode
Observed Behavior
Messages from daily cron jobs were delayed by 3+ hours:
| Event | Timestamp (UTC) |
|---|---|
| FX100 cron job enqueued message | 2026-02-20 04:26:36 |
| Message actually delivered to Telegram | 2026-02-20 07:48:25 |
| Delay | 3 hours 22 minutes |
Logs
2026-02-20T07:47:59.941Z: Found 44 pending delivery entries — starting recovery
2026-02-20T07:48:25.166Z: Recovered delivery f085a33e... to telegram
2026-02-20T07:48:50.322Z: Recovered delivery cd138bd0... to telegram
2026-02-20T07:48:50.326Z: Recovery time budget exceeded — 42 entries deferred to next restart
Root Cause
The delivery queue processor appears to only run during gateway startup/recovery, not as a continuous background process. Additionally, it has a strict time budget that only allowed processing 2 of 44 queued messages before deferring the rest.
Expected Behavior
- Delivery queue should be processed continuously in the background
- No multi-hour delays for message delivery
- Or: Configurable delivery retry intervals
Workaround
Restarting the gateway forces delivery recovery:
openclaw gateway restartImpact
- Critical for cron-based notifications (daily reports, alerts)
- 40+ messages can backlog for hours
- User receives stale information