Notes from our IRC discussion today: 1) The OOME needs to be treated as an error. When the server reaches its limits it should degrade more or less gracefully, but not fail. 2) The OOME occurs when your tests get the load up to levels that aren't attainable with 4.0.5. Thus the absence of the problem in 4.0.5 + JBC 1.4.1 + JG 2.4.1 doesn't really say all that much. Good thing though is this is likely not a regression. 3) You're going to try to confirm the absence of the problem with JDK 6. Before you didn't have enough runs to say for sure it doesn't appear there. You'll also trying setting UDP.use_outgoing_packet_handler="true". View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4032281#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...