Yes the test fails over but no messages are sent to the live server after failover. This is what happens.
a cluster nodes 0 and 1. 0 has a backup of 2. and 0 is the master node that holds the grouping info.
send messages to node 1 which are pinned to a group, grouping info on node 0.
force node 1's cluster connection to failover to node 2.
send messages to node 2 and check they are pinned to the same queue.
actually you are correct, i should be sending the second batch of messages to node 2 not node 1. I will fix and test