Math Overflow is the only SE site I know of that is specifically billed as for the “research-level”.
I don’t have good ability to predict why some suggested sites take off and others don’t. (I tried to start one for translation technology before once, and it didn’t gather any momentum).
Using Math Overflow as an exact template, to clone, what say we explore the feasibility of a research-level linguistics SE - “Linguistics Overflow”?
Supporters?
I think SE is, by design, not at all suited to research-level questions. The idea of SE is that you can ask a quick fact question and get the correct answer. For example, "how do you define an HTML style that adds fixed text to the beginning of a span?". It is not designed for conversational exchanges.
In real linguistic research, there aren't many simple fact-oriented "answer and move on" questions. A question like "what is 'contrast'?" would at the actual level of an actual research site engender thousands of pages of answers, and is still way too broad.
I should also point out that the evaluative infrastructure (anonymous unvetted up- and down-votes) is contrary to the ideal method of academic research, although perhaps it's good if people get accustomed to the popularity-based theory of intellectual value (cf. RateMyProfessor). The problem is that people who hold unpopular beliefs have a strong disincentive to participate in a basically popularity-based website. Popularity is not the same as truth, and I think it would be important to first make clear what you are looking for, and whether you are likely to actually get it using a particular tool.
A well-structured threaded "conversational" forum is a more likely internet-based means of getting the kind of discussion that you seem to be looking for,
I believe MathOverflow is needed mostly because the main Math site is so busy, mostly with questions from students. This site gets questions at a much lower rate.
I would be happy to be proven wrong, though.
