Scenario:
1) Two Artemis servers are running in cluster
2) On cluster connection is set <reconnect-attempts>1</reconnect-attempts>
3) Messages are sent to input queue on Node 1 and read from Node 2
4) Network failure is executed
5) Cluster fails to reconnect, one reconnect attempt is not enough because of network failure duration
6) We wait some time until all messages (2 minutes) from Node 2 are consumed and then shut it down
7) Then we restart Node 1 and consume all messages which was not transferred to Node 2 from input queue.

Expected:
Number of sent messages should be equal to sum of consumed messages from Node 1 and Node 2.
All undelivered messages between cluster nodes should be immediately transferred back to input queue when maximum reconnect-attempts is reached and and connection is not established.

Actual:
Number of sent messages doesn't equal to sum of consumed messages from Node 1 and Node 2. In queue which was created for clustering purpose are undelivered messages, waiting for cluster to reconnect.