Let's say that a journal article submission wordcount limit is 10,000 words. Does this mean that only really awesome or important submissions should be over 9,500 words? Or any article can be at 9,999 without predisposing the editors against it?
If you have 9,000 words, do not try to add new words only to get closer to the limit.
If you have 9,999 words, do not remove words merely to get further from the limit.
Submit no more than 10,000 words unless you ask permission (sometimes there is flexibility in the word count, sometimes not).
In summary, use the limit as given and don't concern yourself with gaming the system beyond that.
These sorts of strategies and games might have a place in standardized testing but they don't have much place in research.
I'm sure someone may say, "depends on the paper" or "the discipline". However, generally speaking, these limits are guidelines to prevent the submission and processing of articles that just do not fit the parameters of the journal. So don't go over (without permission, there may be a bit of leeway).
However, no editor is going to say, "this could be a good article, but it's too close to the limit, so I spike it." If it's potentially publishable, it will go to peer review. And then be treated per the results of the peer review.
If your paper is too wordy for its content and importance, the peer reviews may be less good and the editor is more likely to pass on it. You should also expect peer reviews may point to sections which are too detailed or superfluous, or -- more likely -- highlight areas you should augment. You will then have to fight how to create space for the revisions requested, what to cut out, or plead for extra space. So by and large, if a journal's limit is n words, I would hesitate to try to place an article there where my first submitted draft was more than 0.9 n, say, just because I expect difficulties further on. And I would be doubly careful if I ended up at 9,999 vs a limit of 10,000 after careful chiseling already, since I've never had a minor revisions or revise-and-resubmit that didn't end up growing the paper at least a bit.
I'd be a bit more careful with journals that say "typical papers are n words, but exceptionally up to m > n words". They are telegraphing that their limits are flexible for the right paper, but it had better be something special if you're at m-1 with your submission. That being said, exceptions can always be made. My spouse, not in my field, has anchored an author team that got a highly important systematic review published in a special issue all on its own of a fairly important journal since the paper's required length to be comprehensive required more pages than the journal's typical single issue, never mind single paper. And I've seen important research papers in my old field, math, that have been split into 2 or 3 somewhat arbitrarily chosen subarticles due to the length, and then all published in a specific top journal. But that's rare.