I'm writing this answer as a (pseudo)-grad student who is comfortable using the site, having no real publications or positions under my belt.
Disclaimer: Generally I try to avoid using the pronoun 'I' excessively in MO posts because it distracts from the mathematical content of the post, but I will be abusing it in this post and writing in a more informal style to emphasize that these are just my opinions, man.
First off, I have immense respect for the communities I've interacted with on this site -- initially I pestered the set theory and model theory communities and more recently the category theory bunch, and both groups are rife with experts on the cutting edge of their field and include several heavy-hitting names I couldn't interact with directly anywhere else (except maybe at a conference, but I've never been to a conference and social distancing and all that.)
I wouldn't be nearly where I am today as a mathematician without their contributions to my questions and this site in general. All said and done, I think MO falls easily on the 'good things' list in life largely because of the contributions of experts like the ones I am about to constructively criticize, so please take the following commentary in the least insulting light possible.
The talk of 'gate-keeping' in jdc's post struck a nerve with me, and I think there is a good amount of truth to it. The fact that I like this site has heavily to do with the fact that I am a masochist; I only play video games on the hardest mode available (even if I end up breaking controllers and destroying whole systems), I purposefully don't prepare for lectures or talks to force myself to come up with them on the spot, and I post to MO knowing that I might get a dunce cap strapped to my head while someone tries to make me read hieroglyphs out loud to the world.
I think this sort of 'trial by fire' is one of the best ways for a person to quickly up their game, and MO is a paragon for this sort of mental honing stone in mathematics. I can't count on both hands the number of times I've been embarrassed because I said something dumb to an expert on MO, but I also can't count on both hands and feet the number of times I've had my own knowledge dramatically improved by these experts responses to my dumb questions. I do get the occasional "f*** off and read a book" vibe from the communities I interact with here, but sometimes that is appropriate -- MO is meant to be a repository for research level questions, and if the answer to a question is non-specific (not a single lemma) and covered at large by a well-known text it is appropriate for the questioner to leave and do some reading before they come back to ask a more specific question.
The only type of interaction I've had on here that is totally valueless is the dreaded 'anonymous downvote with no explanation', in particular if it is within the first 10-20 views on a question that isn't obviously stupid or too low-level. I had this happen a few times early on, and if I were the sort of person to take discouragement sitting down it may have dissuaded me from further participation; fortunately for me haters are motivators, so these anonymous and useless interactions only drove me to improve my questions further and make sure the community at large saw the value in them. I think that the dunce-capping and hieroglyph-reciting mentioned above, while unpleasant, serve an immediate and beneficial purpose for the site and to the users of the site -- the anonymous downvotes with no explanation do neither, and stand to cause harm to the sites reputation and the psychology of its newer users. I thusly propose that we collectively cut that s*** out, and that we collectively endeavor to make the dunce caps as small as possible when they are needed.
There is something to be said for keeping the front page neat and free of homework or lazy questions, but I think the flag/vote to close system does a good job of handling those issues. It was mentioned recently in the discussions on these questions that there was an old, informal policy of giving one downvote to obviously misplaced questions so they wouldn't appear on the home page*, then leaving them at one downvote until closure. I support a return to this informal policy, along with a short comment explaining that the downvote was given for this purpose and a recommendation for where to take the question, whether that be MSE, a professor, or where the sun don't shine (for the occasional advertiser.)
*Martin clarified in the comments below that the threshold for a question not being visible is $-4$, not $-1$.