With azure-eventhub on pyamqp, long idle periods (over an hour) can end with the link detached ('connection was inactive for more than the allowed 300000 milliseconds') and the consumer silently stops receiving events, which also floods the metrics blade with User Errors. Raising idle_timeout doesn't fix it, it just delays the failure. The reporter's workaround was to switch back to uamqp for the consumer client (add uamqp to dependencies and use it as the transport) until the pyamqp-side fix lands. If your async EventHub consumers go quiet after idle periods on pyamqp, try uamqp as the transport.
_Source: gh:Azure/azure-sdk-for-python#28996 (Azure Key Vault)_