Since very early in Puzzling's development, we've adopted a very strong principle that's become virtually universal: "if there is a hint of a notion that a puzzle might be good, we have to act as though it is." This is also not a principle that I think has been explicitly or clearly discussed. (It's also not universally true - but functionally speaking, it's pretty dang close. The only consistent exception I've seen is when someone posts a really bad code puzzle.)
However, it most certainly has come up, and often. We very frequently vote this way - but beyond that, it very frequently stops us from spotting and potentially handling low quality content. This goes further, though: it directly blends with another critical notion: "you can't know the quality of a question until it is answered."
But here's the problem: If you can't know the quality of a question until it's answered, then we always have to act as though there's a chance the question might be good. But if we assume there's a hint that the question is good, we have to act as though it is. So in all cases, we have to act as though the question is good... unless it's obviously not, at which point we downvote (or more frequently, don't vote at all), and feel super uncomfortable about our dissonant selves.
This deadlock leads to a voting culture where bad questions are frequently upvoted because they might be good, rather than because they are; it leads to a set of close reasons that must be applied retroactively; it leads to a system where we are optimizing for sand, not pearls. We're stuck trying to resolve the difference between two incompatible ideals: the principle above, and site quality.
The tools we're given on Stack Exchange work best when they're applied before a question is answered. There's a window of opportunity between when a question is posted and when it starts to generate responses, and past that window, what we do has extremely diminishing returns.
I've been pondering this issue very frequently recently, because it seems that at every turn, when anyone proposes something to improve site quality on Puzzling, we run facefirst into this deadlock.
I want to open a discussion on the way we talk about site standards, because I think the way we're doing it right now is ultimately self-defeating. The reality is, we have a lot of information on what makes a good question. We're graduated - and besides, take a look at our active users page.
So let's be honest: often, we really can read a question, and know if it's a good one, without knowing the answer. We should, at this point, be able to say "sure, this question might be good, but nearly all similar cases I remember have been bad." I want these cases to feel actionable, either through votes, or close votes. And in my honest opinion, that can't happen without changing the way we think and speak about the quality of puzzles we see here.
I'm already rambling on too long, but this is pretty important. I'd love for this to be open to discussion - including, and especially, disagreement - because at minimum, I think the discussion will be helpful going forward.