While that may be true, I don't think the kernel hackers consider the Linux kernel to be a short term hack any longer, Linus' original release announcement nonwithstanding. The kernel thus far has survived without copyright assignment, and indeed many consider that a feature.
My understanding is that the FSF wants copyright assignment so that it can unilaterally determine its destiny. It can, for example, decide to begin releasing it under an entirely new license at some point. While it won't eradicate older versions under older licenses, it does make it easy to relicense the whole thing even if the original author didn't submit code with a "version X or later" clause.
I seem to recall that the FSF also argues that it makes it easier to litigate on behalf of the code, although I would counterargue that the critical mass of code that's already there is likely sufficient for litigation purposes. I guess the only sticky points would come down to litigation around a specific contribution that hasn't been assigned to the FSF. I suppose with all the lawsuits out there, that's not outside the realm of possibility. And, to further poke holes in my reasoning, more recent code is more likely to trigger lawsuits than code that's been there for many years, especially if it's based on newer techniques. Older code is more likely to have been vetted against legal issues.