If the node where your Messenger system is running goes down, it fails over to another node in the cluster. Messenger clients reconnect automatically as soon as the Messaging Agent restarts on the next node. Users who are actively carrying on conversations notice the interruption, but do not need to do anything to reestablish their conversation when the Messaging Agent is up and running again.
In comparison to failover, migration typically takes longer because the Messaging Agent methodically terminates its thread as part of its normal shutdown procedure.