Questions about papers in arxiv.org which represent learning opportunities are now discouraged and flagged as improper and closed.
As one of the close voters of the question you linked to as an example, note the issue for me was instead that the OP was basically asking us to try to refute or verify an online paper written by an apparently different person. To me, that question didn't have much context (e.g., the OP did not make much, if any effort, to check the paper themselves and was not specific about anything which may be wrong with it), with it being somewhat similar to a PSQ (Problem Statement Question). To the best of my knowledge, these types of questions have never usually been acceptable on this site. Also, I've read several MathOverflow meta posts and various comments which indicate they are also generally not acceptable on that site either. An example, from just about a day ago, is LSpice's question comment of:
It looks like you may be asking for help checking or completing a proof. In general, MO is for asking specific, focussed questions with a well defined answer, not for checking or completing a research project.
If your referenced post of Status of the Nichols claim to proof of the Collatz conjecture? had, instead, been a more "specific, focused" question, such as with it including something like
In section $1.2$, I don't understand how equation $3$, i.e., $$\text{<Equation }3\text{ expression>}$$ is valid. Due to condition $A$, then doesn't the $B$ theorem apply here so that equation doesn't even converge?
then, instead of voting to close, I would have likely instead done at least one of: just bypass the question, upvote the question, check on the issue in the linked paper, write an answer explaining how the concern is valid or doesn't apply and/or vote to reopen the question if it were closed.
Several months ago, there was a video on a popular series (I believe it was this Veritasium one suggested in soupless's comment) about the Collatz conjecture. This site then got quite a few posts likely stemming from this, e.g., of people who claimed they have solved the conjecture and asking us to verify their proofs. I got the impression from various question comments, the CURED chat room, etc., that this annoyed some members, with anything to do with Collatz then becoming somewhat taboo. Also, there were some related comments in the chat room expressing concerns about the use of the proof-verification and solution-verification tags.
Sometime later, I believe it was around late November, you posted a question asking about your attempt to prove a quite restricted version of the Collatz conjecture you had come up with. Your question explained the background techniques you were using, what the variables were and stated your proof attempt relatively well (e.g., quite directly and succinctly). You also indicated where you thought there might be an error, with a question comment indicating a mistake (I believe it was in the part you suggested), and with you confirming it was an error.
As far as I could tell, your question met the various conditions mentioned in the How to ask a good question. post, with many other rather similar questions being asked here, but with few of them being closed and many being answered. I was somewhat surprised your question was closed, with it now being deleted. The Collatz conjecture association may have been a factor, but you could also perhaps have improved your question by, for example, explaining in more detail about why you weren't confident about the one step you mentioned.
From my perspective, there has not been any overall change, at least relatively officially, in the acceptability of questions about speculative mathematics over the past few years, as long as they are otherwise good questions, such as having appropriate context (if I'm mistaken, I hope somebody, in particular a diamond moderator, will explain any such change, e.g., by writing an answer here). Nonetheless, regarding sites where such questions might potentially be more welcomed, I don't know of any offhand.
IMHO, you're the type of conscientious member we encourage to join and contribute. Note I have found your questions to be generally quite well written, with some of them being among the most interesting (and sometimes challenging) ones I've read here. Regardless of whatever extent you continue posting here or somewhere else instead, I hope you get high quality, constructive feedback so you can continue to learn and grow your mathematical skill & knowledge.