Sometimes happens that jms client is not able to read all messages after failover

There is issue with jms consumer which is not able (sometimes) to read all messages after failover. Consumer just hangs and is not receiving new messages.

Test scenario:

1. Start server A and B in collocated topology - servers contain backup for each other
2. Start producer and consumer connected to server A - clients are active for the whole duration of the test
3. Server A is killed -> clients failover to backup on server B
4. Wait until all messages are read

I've attached NetBeans project with jms client which is used in reproducer. Consumer acks every 1000th message.

How to reproduce:

1. copy following zip from messaging lab and then unzip:
scp messaging-05:/home/hudson/users-tmp/mnovak/tmp/reproducer.zip .
2. start server A in a new console - sh start-server1.sh 127.0.0.1
3. start server B in a new console - sh start-server2.sh 127.0.0.2
4. start producer in a new console - sh start_producer.sh 127.0.0.1
5. start consumer in a new console - sh start_consumer.sh 127.0.0.1
6. kill server A using - kill -9 ... (jps -m - shows processes)