non-durable queues are created only on the live server and not on the backup server (wether using replication or shared store, non-durable queues are not stored in any way).
If a client creates a non-durable queue and fails over to the backup, failover will fail to create the consumer since the queue does not exist on the backup.
the tests I'm adding back for cluster with backups uses non-durable queues.
the client consumers fails to consume messages sent after failover since the consumer does not exist on the backup server...
I could use durable queues for the tests (like for other failover tests...) but it seems to be a valid use case to have failover even when using non-durable queues.
wdyt?
