Limitations of HA

Client level transactions do not span across servers in the Enterprise Server when running in shared mode. Transacted sessions involving receivers will be rolled back if the primary server crashes. Therefore, the messages delivered in that transaction are redelivered to the receivers upon connection to the backup server.
Distributed transactions, which are in execution during the transition phase, become "in-doubt transactions" when the primary server goes down. These transactions get 'rolled back' and can be recovered after the client reconnects to the secondary server.
JMS Topic Requestor can not receive its intended reply if failover occurs after a request is sent. This occurs because JMS Topic Requestor creates a non-durable subscriber, which can miss a message during failover. However, if a topic requestor creates a durable subscriber to listen for replies, then it works even during failover.
If both HA servers (primary as well as backup) go down, the requestor receives a duplicate reply (with reDelivered Flag = true) for the first request made after failover occurs.