Your Postgres logs show it shut down cleanly at 22:17:38 UTC on May 19 (completed checkpoint, confirmed "database system is shut down"), so a stale postmaster.pid is not the cause. The restart failures are due to a container-level error where the runtime cannot locate the process binary, which is consistent with image registry disruption during the May 20 service disruption. Triggering a fresh redeploy from the service's 3-dot menu should pull the image cleanly and resolve the startup failure. Since the database shut down gracefully, your volume data should be intact.