Relevant to this question are the remarks of Grothendieck on "speculation":
I agree with him on the general negative view of speculation. I have tried out some speculative ideas, and you can get some funny looks from superior persons, who can be easily inclined to use the word "nonsense" or "ridiculous"! I have had this with regard to my higher dimensional ideas. Or even over the years with regard to the use of groupoids at all. On the other hand, by giving a speculative talk in 1975 I got a very helpful idea from a later discussion.
On the other hand, I have found talking about these ideas has helped to make them real, and suggest they should be pursued, even how they should be pursued. If someone says "That can't work because ..." then that is useful information. So is "If your idea was any good, it should do .....". That could focus the mind; what would it need to do that? This turned out to be a key!
I would advise writing for yourself as speculatively as you can, taking ideas as far or beyond as you can, trying to catch ideas before they vanish. Then they should be looked at in the cold light of day, looking carefully for obstructions to them working. If these obstructions exist, that would be interesting. If they can be gradually be overcome, that is even more interesting! In either way, you may be doing something different. I expect we all want to find something new rather than solving already formulated problems. How to go about that, and how much effort to put into it, especially for career purposes, is worth discussion, especially in terms of balance of effort and probabilities of success, given one has to produce papers.
There are also dangers about talking of new ideas on a famous problem: I have been told of one well known example of this (elementary proof of the Prime Number Theorem). Unfortunately, there is as far as I know no systematic discussion of such ethical issues, and of attention to due scholarship.
Hope that helps.