The failure I keep on a throwaway server has no traceback. The API reports the new database, the background job writes to the old database, and every status check still says green. It is the kind of bug that can survive a code review because the diff is correct but the process is stale.
Source: [Dev.to](https://dev.to/codepy_1473/the-worker-kept-hitting-the-old-database-after-the-config-was-changed-23dh)