I'm struggling to understand the rationale for radically changing the training system - at least to that degree - and I *strongly* disagree that subspecialists invariably lose or don't have to use their generalist training. Your nephrologist example is particularly poor - certainly at my (Canadian) centre, IM subspecialists routinely cover general medicine call, and the nephrology group in particular tends to be pretty good. I'd go further and say that there is absolutely such a thing as excessive specialization, and the concept of a nephrologist who couldn't (or hadn't ever) managed straightforward sepsis is actually rather frightening. There's also a big difference between, say, a new-ish GI doc who doesn't know how to manage DKA comfortably because he never did an IM residency and a late career CT surgeon who simply hasn't done an LOA for an SBO in 30 years.
The other side of things is that certification standards give rise to a fairly natural monopoly of, well, training standards. A GI group may or may not want someone who is ERCP-trained, but they certainly will expect a standard level of endoscopic competence, to say nothing of an appropriate knowledge base.
In terms of the actual residency training process, your "free market" approach seems to go much too far in the other direction. It actually sounds unworkable, and if anything would require far greater resources to implement. A decentralized system will require decentralized preceptors and evaluation methods, and there is potential for even more heterogeneity and lack of compatibility than already exists. Is this really feasible? We employ centralization and standardization to avoid the
costs of ending up with very unqualified people in practice and financial impact of a decentralized, disorganized training system .
Now, in Canada, the Royal College is embarking on residency training reform to move toward "competency-based" methods:
http://www.royalcollege.ca/portal/page/portal/rc/common/documents/educational_initiatives/cbme.pdf
The implication is that residency training would no longer be of a fixed length and could be potentially shorter or, potentially, longer, depending on the trainee's progress. I am not sure of the implications for flexibility in changing specialties during or after residency, but it appears that it might be somewhat more flexible. Having said that, I favour returning to our former system prior to the mid-90s where most new MD grads completed rotating internships, qualified for general licensure, and subsequently practiced as GPs, started a specialty residency, or worked for a few years and returned later to become, say, an anesthetist. At that time it was common for people to start in one program and end in another; the overall effect is that we had far more generalists.