No, I generally do not. I just always try to clearly define units of work within a conversation so that if possible I end the conversation when the user is "done" with an activity (which usually means wizard-style approaches). I also use propagation="none" on navigation as the user may back button to the conversation at any time to resume execution. The conversation may timeout but I do try to give the users some time to continue the conversation if they accidentally navigated away from some activity. Sorry for hijacking the thread, but I am curious why the issue is any different with manual flush mode? I would think maintaining a detached conversation context instance and merging on submit accomplishes the same thing (except that lazy loading would not be possible). Maybe you are referring to if you merge on every action (which I guess would always flush without manual). Could you shed some light on that? Thanks! View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4086896#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...