I ran the JPA demo 100 times and looked for memory leaks. We are currently leaking storage in SFSBPCMap. This leak will be less once we have EJB3 support for @Remove and the SFSBDestroyInterceptor is invoked.

The remaining issue is whether SFSBDestroyInterceptor will be invoked for SFSBs that experience system exceptions after injection has occurred.