Reliability

Not every update is a new feature.

The pipeline had been running well, but we found failure modes that only show up under sustained load or after long uptimes. We fixed them.

Silent connection failures — where the connection appears alive but has stopped receiving data — now trigger automatic reconnection. This was causing occasional data gaps that were hard to notice but real.

Services now reset cleanly on restart. A previous bug left them in a degraded state after a stop/start cycle where events were processed but not emitted.

Error tracking is now consistent across all services. When something goes wrong, we know immediately.

Memory retention limits added across the board. The system stays lean over time instead of accumulating indefinitely.

None of this is visible externally. But it's the difference between a system that works most of the time and one that works all of the time.