Gurvinderpal,
We're experiencing the "stuck in delivering" issue also, though on a clustered setup. It occurs on queues consumed by MDBs and by JBoss ESB. We don't yet have a solution, but I'm closing-in on one. I'm wodering if it isn't a remoting timeout issue when the server is under high load (we're doing load testing) or during a stop-the-world GC. This poster has had success increasing the remoting timeouts: http://community.jboss.org/message/442276#442276
The next time you notice this, check the consumer count on the queue. Is it the expected value, or is it zero? (I believe MDB-consumed queues should always have one consumer.) What about the delivering count? Does it match the number of stuck messages? What happens if you stop and start the MDB consuming that queue via the jmx-console? The delivering messages will disappear, but do the messages actually get consumed? (have they perhaps already been consumed? Can you prove this via the logs?) The more evidence and information we gather about this, the closer we get to a solution. However, most of JBoss' attention has been turned toward HornetQ (according to my limited powers of observation). Even if you find the exact source of the problem, a solution may not arrive quickly or at all, unless you've paid-up your JBoss credits.