At the end of the tests, there is about 900 PostOffice instances hanging around. I will run some tests later this weekend trying to identify the places that are probably holding the instances. It would be better for this diagnose if the Pingrunnable was shutdown right after the close call... It would be better to identify when the leak happened opposed to a regular delayed shutdown. On the InVM transport, the PintRunnable will indirectly hold an instance to the Server. (somewhere through InVMConnection). I will post more details later this weekend. View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4249757#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...