In some cases, Seam page rendering can fail because of the Conversation locking system and a ConcurrentRequetsTimeout exception is visible in logs.
The root of the problem is that the way Seam handles the conversation concurrency is not ideal.
We can surely do better :
- try to choose the victim : kill the old request and not the new one
- create a temporary conversation that has a change to run : re-init a minimal context
- reduce critical section size