David Henderson talks about it here.
Let's be clear about three things that would still exist if we did away with residency requirements tomorrow: the M.D., the AMA, and (for now at least) private HMOs. Even if you don't have private HMOs public insurance serves about the same purpose in what I'm about to say (I know that may bother some people, but I'm not interested in talking about that facet of health care right now - I anticipate private and public insurance would do about the same thing in this situation).
If I were an HMO I'd differentiate strongly between an M.D. that the AMA approves of and one that it did not approve of. If I were a patient I would too, but let's be honest: often HMO trumps patient. If I were an HMO, I might not cover non AMA approved doctors at all or if I did I'd cover a lot less. Over time we'd learn (this is the essential thing that spontaneous orders do better than planned orders - not that planned orders don't learn, but they aren't as quick to) exactly who needs what kind of scrutiny. HMOs would probably never pay for a non-AMA approved surgeon but they may pay for pediatricians that have intermediate approval levels. We see a lot of efficient division of labor now between doctors and nurses, but you'll see more of that.
One thing I don't think you'll see - so long as we have a substantial third party payer presence - is a free for all where the AMA does not play a role in very closely monitoring the skill sets of doctors. But this would be a situation of certification rather than licensing.
If the AMA was too restrictive in handing out these certifications (not an odd thing to anticipate) you would probably slowly see alternatives developed. HMOs might get into the certification business themselves (what better way to eliminate the information asymmetry?). Even if the AMA was still too restrictive, you would have ways of solving the current problem of too few doctors without resorting to unqualified doctors (although some people would, and we'd have to think about what to think about that).
I strongly prefer this certifications regime over a mandated licensing regime, with a few quibbles. I don't think it's especially illiberal to think there's a public interest in ensuring that no one - no matter what dire straits they're in - goes to a quack surgeon. That sort of thing might justify a certain type of licensing regime for certain types of doctors with a mechanism for keeping an adequate supply of surgeons. But even if we justify a certain type of licensing like this, two facts are still pretty clear:
1. The cases where licensing may be preferred to certification are few, and
2. We still need to worry about adequate supply where we might like licensing
In any case, the system needs to be dynamic. Medicine is getting more capital-intensive which may presage a deskilling of the profession (at least in certain areas). It's bad enough that licensure limits supply and provides unnecessary rents to doctors. At least we get pretty damn good doctors in that bargain. If we still had these sorts of supply restrictions after medicine was largely automated we'd be paying a lot of the same kinds of costs for little benefits at all.