Our Postgres service (basic n8n + Postgres template, Hobby plan) repeatedly hangs in a way Railway cannot detect or recover from, taking down a production n8n instance (message pipeline for a veterinary practice).

Symptoms: Postgres deploy logs simply stop; the process does not crash and the service stays "Online" in the dashboard. n8n then returns HTTP 503 "Database is not ready!" and logs "Timed out after 30000ms waiting for database connection recovery". CPU and memory are flat/low at each hang, so it is not resource exhaustion. Recovery logs show pgBackRest backup/restore-gate activity around the hangs, which may be a factor.

Occurrences (AEST, 2026): Jul 2 ~17:36 (down ~2 days), Jul 10 ~02:34 (down ~34 hrs), Jul 18 ~06:00 (first restart attempt failed), Jul 19 ~08:00, and Jul 19 21:14 (down ~22 hrs, re-hung only ~3 hrs after the previous fix).

Why we can't self-heal: the Restart policy "On Failure" never fires because the process hangs without exiting, and Healthcheck Path only gates new deploys. Our only workaround is manually restarting the Postgres deployment, then the n8n deployment.

Questions: 1) Is this a known issue with the Postgres template (or its pgBackRest sidecar) on the Hobby plan? 2) Can you see anything on your side (host events, volume I/O stalls, backup jobs) correlated with the timestamps above? 3) Is there any supported way to auto-restart a service that hangs without crashing (runtime liveness check)? 4) If this is a platform-level quirk, what is the recommended fix - different template, pinned image, or plan change?

Project app URL: n8n-production-fc89c.up.railway.app. Happy to provide service IDs or logs.