I have succeeded in creating a stand-alone example of the behavior in my original post. This download is a ready to go EAR with source included. Simply drop it in the deploy directory of a JBM configuration (mine was built from `all`) to run. http://www.mediafire.com/?2y3w3gy2xic This example has two producers, one sending priority 4s into a queue at a rate of 50 messages per second and the other priority 9s at a rate of one every five seconds. An MDB consumes from the queue at a rate of 15 messages per second. Watch (you may have to look at the logs afterwards) how the high priority messages, represented by the letter 'H', are sent into the queue but not processed in a timely manner. Use jboss.j2ee:ear=test-prio.ear,jar=test-prio-mdb.jar,name=BootstrapImpl,service=EJB3,type=ManagementInterface in your JMX console to start and stop the senders. You can let them run for a minute or so and then stop them to watch the MDB clear the queue; the 'H' messages appear only intermittently in between large numbers of 'L' messages, even after the producers are stopped. I will investigate the pure JMS solution as per your suggestion, but my application was planned, designed, implemented, tested, and placed into production using MDBs -- whatever the outcome, it is the MDB stuff I ultimately have to get working. I would be enormously appreciative of any further insight you may have. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4126265#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...