If an ACK fails due to a persistent store error during the remove(),
e.g. database is down, the message should be NACKed and an error reported.

Currently, you have to recover the queue (e.g. redeploy the queue)
otherwise the message is stuck in limbo.