We have a webspeed application which supports a small number of users and where the session restoration carries a heavy penalty, so we are looking at how it performs when running state-aware. We are able to lock the agent by executing a setwebstate call correctly with setting the STATE_AWARE_ENABLED environment variable and this seems to work well.
Unfortunately we have noticed that when in the locked state, the user clicking a form post is only allocted their locked agent if it is idle. Once it is given something to do and if the user does subsequenct clicks which result in form posts that the ws broker allocates the request to an available agent and not our locked agent.
The session then acts like its state-aware (by seeing the wseu cookie) but then thinks that the state-aware procedure dos not exist as it cannot find it. This is due to the fact that the code is suddenly executing against the wrong agent. It then deletes the cookie and the application dies a death.
Any ideas on how to work around this. We haven't done uch in this area. I think its a bug but am wondering if we are missing something.