Most of my questions on math.SE come up in the course of personal investigations into some topic, and usually get asked after a fair bit of reflection and attempted work on the problem. Unlike questions from a textbook exercise, I have no idea a priori what sort of effort might be required to solve the question, and states of partial progress are very common.
As such, I'm frequently able to make substantial but less-than-complete progress on a question, and when I post, I'm looking for ways to extend my existing work to a full answer.
In such cases, it seems to me that there are two options:
Post a (rather long) question, giving both the problem at hand and a summary of my work thus far, with enough detail that relevant strategies and papers and such could be utilized by future answerers of the question.
Post the question alone, and then post a self-answer, giving whatever progress I have separately and leaving the option for future answers to fully resolve the question.
As an example of such a situation, take this question. I realized early on that there were different strengths of restriction I could impose on the problem, and as I found examples for multiple cases I included them in the draft of my question, but by the end the resulting post was quite long and involved, and parts of it feel rather more like an answer. On the other hand, with this question I posted an improved bound to the question as an answer rather than editing it in, given the separation in time between the two, and perhaps it should have been an edit to the question instead?
I can see some points for and against both options:
When parts of the content are written in a much more definitive and proof-based way, it makes sense to group them with the answers, for better separation of the relevant parts of the question and to allow users to address the two pieces separately in two different comment sections.
Posting the same content across two answers could be seen as a karma-farming attempt, by trying to get upvotes from users on both of the submissions at once, and should therefore be avoided.
Having the question show up on feeds as already answered could reduce its visibility (even if it is not accepted), because people assume that the question has already been wrapped up and so don't click on it in the first place. (Because the answer is not complete, unlike most self-answers, I want for the question to still get viewed by potential solvers.)
It is not clear to me how I ought to balance these considerations, and when to take each approach. Any guidance would be appreciated!
This meta post seems a bit related, though it deals with a rather different scenario.