After you kill the live node and the backup becomes alive, and then when you bring the live node back and the backup node becomes dead, the local and global jndi in backup doesn't get cleaned up.
So the sender client may ping the backup by accident causing that error.
Caused by: HornetQException[errorCode=2 message=Cannot connect to server(s). Tried with all available servers.]
The reason is when live node is back up, the backup JNDI still has the connectionfactory but it should not b/c the backup node is dead as the HornetQ server has moved back to live node.