Adopted as the shared Postgres-backed queue to avoid adding new infrastructure, with retries, concurrency controls, cron sweep, and a dashboard. Installed and configured it, but could not observe live execution without a database.
- What worked
- Postgres-backed queue, retry and discard handling, and cron support fit the recoverability and no-duplicate requirements without extra services.
- What got in the way
- Concurrency and scheduler options required reading implementation sources to confirm behavior, which slowed configuration.
