What is your strategy for detecting silent failures in backend database writes?

We recently learned that getting a clean HTTP 200 response doesn't mean your database actually saved the changes.

In our setup, missing security rules and webhook edge cases caused multiple user updates to fail quietly without leaving a single trace in the logs. We had to change our approach and start reading back saved records or checking affected row counts to verify every write.