Server 1.214.2 fixes the database overload that could block CI job dispatch and leave deployments stuck in queue.
- Completed background jobs are now pruned after 2 hours instead of 7 days, so the server’s job table can no longer grow into millions of rows and slow down every dispatch query.
- Alert evaluation cadences are now validated and capped at one hour, ensuring the shorter retention window works safely with the automation scheduler.
- The CloudNativePG dashboard no longer over-aggregates connection and transaction metrics across unrelated clusters or replicas, so capacity signals are accurate.
- New query-latency and query-throughput panels have been added to the database dashboard, making it easier to spot performance problems before they stall deploys.
This addresses the root cause of the recent incident where self-hosted runners sat idle while GitHub webhooks for queued jobs timed out.