I aim to solve this before the GA release.
Hopefully (?) in the next 8 weeks or so, I would guess.
This is critical functionality for us also. I am glad to here that this is in the plans for GA.
Please be sure to include the use case of the failure and restart of the fail-over node; this seems to be often overlooked.
Dynamic group membership with state synchronization is CRITICAL to a robust enterprise queuing solution! Even if the the cluster is only a 2-node active/fail-over pair.
BTW: I really enjoyed your presentation at JavaWorld; but some of us do read the manual first :)