I debugged by scaling down the pods to 1 and replacing Task.Run with await, but the issue persisted. After an hour of troubleshooting, we found the real culprit—an active trigger on the table.

With .NET 8.0, triggers need explicit handling, or they cause silent exceptions logged only at the console, not in DB or files. This also highlighted gaps in our error logging.

So, Task.Run wasn’t the issue.