https://github.com/EnMasseProject/enmasse/issues/5226

For deployments with large numbers of addresses/connections, the volume of stats data may be sufficient to block the NodeJS event loop for excessive periods of time. This means other processing activities may be delayed. This may include the timely processing of Kubernetes probes where a delayed response can lead to spurious probe failures and container restarts.

The issue is the address and connection stats data is processed in one go, without yielding control back to the event loop. In a busy system, say with lots of active connections, there will be significant work generating/handling the update events.