Two things need additional color, one being given the lack of atomosphere at the altitude where the first stage separates (tapes put it at just over 3 minutes into the flight), and two structural integirty of those struts making the atmospheric transition.
So you need to get the retro burn off while you are exo-atmospheric or you risk melting off the important bits when you re-enter. That suggests that you need some sort of RCS to cut rotational dynamics so that you can stage for retro, then re-enter at less than hypersonic velocity. Further you can't do you retro burn until you've allowed the upper stage to clear.
Another interesting bit is that the first stage appears to have cameras, they were showing video from them during the launch, but we've not seen the post separation video. I hope that video surfaces at some point as it could help design something to mitigate the whole roll thing before it got out of hand.
[1] http://www.reddit.com/r/spacex/comments/1ne36t/partial_succe...