Having now had a little experience with the site here, (and almost none with SO), I find the acceptance idea to be less compelling here.
I've been going through peoples old questions and giving better answers than their 'accepted' ones.
Is this sensible? I don't know if it will benefit my reputation here, but a lot of the 'accepted' answers for 1 answer questions are, frankly, shoddy work. Either uninformed, or lazy writing or some combination. Assuming that they will be searched and found in the future, the answers should be better than they currently are.
That said, the nature of these questions and answers are very different than "Is there an O(n log n) solution to Y problem?"
One HUGE difference here as I think about this is it's highly likely that the questioner may be totally unable to validate the correctness of an answer given. That's very different from programming, where essentially our compilers/interpreters validate the correctness of an answer.
I may know exactly why your hollandaise breaks, or your ragu is bad, but that doesn't mean you will be able to fix it based on my advice.
I'm less sure about solutions -- I can think of some possibilities -- abolishing correctness, or just giving a 'correctness' bonus to the highest rated comment, in effect crowd-sourcing the 'correct answer', or possibly 'correct answer' is voted on so that everyone can choose their own correct answer and keep it different from the ratings. Here is where my lack of knowledge of SO hurts -- I'm not sure what analogous situations have come up already elsewhere.
Another possibility is to encourage people to 'try out' answers before accepting, and then you could always mail them periodically to come back and accept. This could be a sort of community solution to the opportunities / problems the software presents.