My take on this whole issue is that there *will* be a stable version of Wikipedia, because there is demand for such a thing. When my mom reads articles, she doesn't want them to sometimes be in the middle of an edit war, or sometimes have been vandalized, and so on. As Wikipedia currently is, on average it has great information, but at any given time, any given article could be completely crap, even if it was great 10 minutes ago. Sure, those of us who know what we're doing can look through the edit history and pull out a good version, but most people aren't going to want to or know how to do this.
So basically I think if we don't do it ourselves, somebody else will start sifting out "good articles" from Wikipedia. I think we can do a better job of it though, and keep the results under the Wikipedia name.
I also don't really see what the worries with elitism and especially with the FDL are. I see this exactly as most open source software development -- you have a CVS branch that can be updated at any time, and periodic releases of "known good" versions. The attributions and such are of course not going to be removed (I wouldn't think).
I think perhaps some of the problem might be that people are advocating (or think other are advocating) a more formalized process along the lines of Nupedia. I don't think that would be a good idea -- choosing "known good" versions of articles should IMO by done by consensus, and we shouldn't require any formal credentials to do so (though hopefully people with such credentials will be among the people who make comments in the attempt to reach a consensus). But I do think some process by which a particular version of an article can be nominated as good, and then added to the "good" distribution if there are no objections, is a good idea.
This could all be done on the Wiki, but if the stable distribution is to be anything but a tiny subset of Wikipedia, I think some more software would be necessary. Some automated method by which the software keeps track of submissions and comments would be helpful. One possible method -- anyone can nominate a particular version of an article, and anyone can post replies to the nomination that are tagged either "support" or "oppose" (or "neutral"). Any nomination with no "oppose" comments within some certain period of time is automatically added to the stable version; the ones with opposing comments are dealt with more manually in the usual wiki way, added if it's been determined a consensus has been reached, or the nomination withdrawn if a consensus against is reached.
Of course this all requires someone with time and interest in coding up some more software. But I think that's necessary -- the "Brillian Prose" manual method doesn't really scale well, and will never I don't think end up successfully scaling up to, say, keeping track of 25,000 articles.
-Mark