Following up my own post. It seems that: - during login(), new UserSession; updated User.userSessions entity and collection aren’t put() into cache by design. (see #4 at the very bottom of the previous post). If my guess is correct, whose responsibility is to evict the existing (stale) user.sessions collection node from cache? - Nothing evicts the empty PersistentSet collection entry already sitting in cache. So, on subsequent read request, an undesirable cache hit takes place. But why isn’t this happening with when testing with 1 user, or for every user when multiple threads are employed? View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4214385#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...