Consider this. It's been building up votes for weeks, but it is most definitely not attracting answers. The comment trail suggests that's it's just a plain old bug in the compiler. None of the commenters seem to be inclined to turn their remarks into an answer, so, there it sits: right at the top of the unanswered list, advertising the apparent inability of the site to answer it.
I think litb put it very well,
If the question absolutely requires all that information in order to answer it, then it's probably going to sit there until an expert comes along, gets interested in it, and solves the problem.
However, I suspect that a several very short questions would more likely reach the correct conclusion.
Further, in the question itself, the author states that his query is merely of a curious nature, he's resolved the issue for his own application:
So only those that
will even consider approaching the problem.
Lastly, I've found that I get much more response if I do a teaser question. It turns out that if you draw people in with a very short, simple question that strikes at the heart of the problem (maybe with a very short (2-4 line) code snippet that shows the problem), then you'll get a ton of people interested enough to ask detail questions. Further, since I'm not sharing my assumptions, they will be very well questioned.
Turns out a lot of my "superhard" questions are merely poor assumptions I can't even see that I'm making.
But by getting people invested in the question, I can then travel down the path with them, even though I've already gone down that path, and they can show me new avenues of solutions, and will spend a lot more time on my problem than they do if I lay it all out for them in a 2 page summary of the issue.
So litb is largely right - too long is a problem here.
It actually has 3 answers, but they are all deleted for various reasons.
Seeing as how the question is highly specific and no one can seem to reproduce the problem, I guess people may be upvoting (possibly out of sympathy) because the question was at least complete and well-written, and potentially useful to someone else who runs aground of the same bug.