I don't see any major issues that would arise with cross-feeding, and find it hard to believe everyone hasn't been doing it for the past 50 years.
I don't think Musk is going to lose much money pursuing at least first-stage re-usability unless he has a lot of crashes during grasshopper testing. I doubt the landing legs and minor enhancements add much to the cost of the stage, so when he loses one it's not much different than just expending what would normally be dumped in the ocean anyway. Upper stage re-usability will of course be trickier.
If he's successful with both first-stage re-usability and the F9 heavy, I think NASA should carefully consider the idea of mating 2 and 3 F9-Heavies together to give them a fully re-usable first stage that would outperform SLS, where six booster stages seperate and return, then three cores. It would be a shorter stack with less risk and very little added development costs, and the components could still be used individually, keeping their flight rate up by filling in with smaller launches.