MSMQ transport can't recover from connectivity issues

[SzymonPobiega]

When the MSMQ service is down, the MSMQ transport signals the critical error (as it should) and logs problems (as it should) but when the connectivity is restored, it starts throwing different exceptions:

System.Messaging.MessageQueueException (0x80004005): An invalid handle was passed to the function.
   at System.Messaging.MessageEnumerator.MoveNext(TimeSpan timeout)
   at NServiceBus.Transport.Msmq.MessagePump.<InnerProcessMessages>d__6.MoveNext()

and the endpoint is unable to resume processing of messages until it is restarted. Combined with the default critical error handling of noop this makes it hard to notice endpoints that are not processing any messages. The expected behavior is for MSMQ to resume processing after the connection is restored.

Symptoms

The endpoint stops processing of the messages until restarted

Who's affeected

All users of the MSMQ transport are affected

Workaround

Define a dependency between the Windows Service hosting the NServiceBus endpoint and the MSMQ service via the Services UI (services.msc). This dependency will make sure to stop and restart the NServiceBus endpoint after MSMQ is back online.

Related issues

Originally raised in the NServiceBus.Router repository