Ok, with help from Andrew on the related thread http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4041276#... I see why this message is appearing. When a callback handler is registered, the client side creates a control connection, and when the handler is unregistered, the control connection is destroyed. However, the creation and destruction is somewhat asymmetric: the connection is created when BisocketClientInvoker.transport() sees an ADDLISTENER message go by, and it is destroyed when the callback BisocketServerInvoker.handleInternalInvocation() gets a REMOVECLIENTLISTENER message. Originally, the treatment was symmetric, but in the case of "quick shutdown" with no network i/o, BisocketClientInvoker is bypassed, which caused a leak. Now, the problem occurs when the Messaging connection is created in a single JVM. Even though the transport is specified as "bisocket", a LocalClientInvoker is used, which makes direct method calls on the server invoker, and so, appropriately, the bisocket control connection isn't created. But the REMOVECLIENTLISTENER message still goes to the callback BisocketServerInvoker, which tries to destroy the non-existent control connection. I'll reduce the message's log level in Remoting 2.2.0.SP3, which I'll create in time for the Messaging 1_2_0_SP2 release. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4041278#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...