The tests work now. I committed the fixes and the complete version of Ales tests. I have done a clean checkout of deployers tests and confirmed they are passing for me. This is a complete list of the Jiras involved with the bugs discussed in this thread: https://jira.jboss.org/jira/browse/JBDEPLOY-216 https://jira.jboss.org/jira/browse/JBREFLECT-65 https://jira.jboss.org/jira/browse/JBREFLECT-66 https://jira.jboss.org/jira/browse/JBREFLECT-67 https://jira.jboss.org/jira/browse/JBREFLECT-68 https://jira.jboss.org/jira/browse/JBREFLECT-69 (the bug responsible for the random failures) Notice that, despite the wide coverage of the class pool tests Kabir wrote, most of the bugs were uncovered by the deployers-vfs tests. The reason for this is that those tests were reregistering stuff, which hid the bugs (JBREFLECT-67). Now, I don't know if we need to port the vfs tests to class pool or not. Maybe I should only port the "key" tests, that uncovered bugs. With the refactoring performed for JBREFLECT-68, I'm thinking on getting rid of RegisterModuleCallback, incorporating its code to VFSCLDomainRegistry (which, BTW, I'll probably rename). View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4266833#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...