Considering that most of the work involved would be XML config files (MC beans) and perhaps an integration class or two, I don't see the need for it to affect the AS source tree (at least not in it's current "tech preview" state).
Agreed, let's not touch the AS source tree where possible. Simply because of the size and complexity of it.
I was hoping to target AS/EAP 5.x first and then move on to 6, but FWIW any such profile for 6 should then make use of Paul's work in deploying a CacheManager. In the case of HotRod and Memcached, setting up an endpoint involves selecting the appropriate ProtocolServer implementation
Again, +1, it makes sense to start there since there is probably a much bigger user-base for 5.x. More people using it, more bugs reported since it's just gonna be a preview. More bugs reported = better versions of this profile come AS/EAP 6.x which I believe is where we expect this profile to be a lot more useful?