Yes, we need to clean things up
For a Q&A site to come across as useful, the chaff simply needs to be separated from the wheat. We can't expect external search engines to care about our internal quality metrics; at best we can get them to care about how many times a post is linked internally, and even that's pretty wonky. If search results for things on Codidact turn up bad questions, that makes the site look bad - even worse than if it fails to turn up relevant questions.
But maybe don't delete things aggressively
It might be better to shuffle things off automatically to a different category - as long as it's done in a way that search engines can meaningfully care about. It's good if people get (correctly) the impression that Codidact communities can offer more individualized help as well as putting forward a high-quality canon, without mixing the two. That's the sort of thing that makes a community feel like a community: being both caring and goal-oriented.
If things that aren't in the main Q&A section can be held to different (lower) standards, that offers the opportunity for a more welcoming experience for everyone. Instead of having to filter tons of junk out of the place where properly-formed Q&A is presented, curators can cherry-pick from an explicit, more-help-desk-like pile to see what issues are common, create canonical Q&A, and then either start closing duplicates or at least providing a reference that answer-writers can use to shorten their answers (dramatically, in most cases). People who have a problem and really want a forum experience (or debugging help) don't have to look elsewhere (although I imagine we would still have considerably stricter requirements than most actual discussion forums), and people who like answering those kinds of questions can take advantage of an environment where the question is fixed in-place at the top of the page.
However, for this to work, it needs to be easy - ideally, even the default - for web searches to land people in the top-quality content. Making this work properly might involve using subdomains carefully.
I've remarked before that it's not very useful that a "needs author's attention" flag simply redirects people to the comment interface. As has been proven to me time and time again in a Specific Outside context along with many other Similar Experiences, there are all kinds of people out there champing at the bit to be "helpful" with barely comprehensible beginner ramblings, making herculean efforts to read OP's mind and interpret a question when nothing has been clearly asked - and the availability of AI has only made it exponentially worse.
In the long run, expecting multiple people to act to close a bad question before anyone can try to offer answers that clog the site and drag down average quality... is simply not scalable. Not even if you have multiple clearly-defined reasons for closure and a whole subsection of Meta dedicated to explaining the how and why of it. By my reading, "needs author's attention" is largely a catch-all for questions that:
are unclear (whether because they are written in a language the community doesn't want to support, are seeking to do something that doesn't make any sense to attempt, don't clearly define a recognizable problem...)
are missing specific crucial pieces of information (this can cause the perception of being "unclear", but is really a separate issue)
are trying to ask too much
reflect an idiosyncratic issue that prevents the question from contributing to a Q&A canon that can help others (whether it's a "typo", logical oversight or whatever is equivalent to that for a given community; or is just otherwise too specific to an individual's exact setup)
are subjective in a way that doesn't work well in the Q&A format
My assessment is:
This is missing the problem of questions simply being blatantly off topic. If I want to flag such a post, it seems like I'm expected to use the "other" reason, and I think that's absurd. Even if it's only subtly off topic, it might easily be recognized as irreparably off topic.
The idiosyncratic issues, on the other hand, are the exact sort of thing that I can imagine being permitted in a separate category. For example, if there are a lot of people who want to debug code, a lot of people who want code debugged, and a lot of potential to "farm" good Q&A out of that... I don't want to keep those people out of the Software community, I just want to keep their "questions" and "answers" separate from the actual Q&A.
Aside from those two points: the key metric for a "bad question" is that it "needs the author's attention" and doesn't get that attention. Aside from blatantly off-topic content, these are the questions I'd like to see actually deleted. Even in an explicit "tech support" context, there's no point in keeping a question around if answering it requires input from one specific person and that person isn't forthcoming.
On the other hand, the lack of such information is a reason to try to keep the question "out of the way" - in a place where curators can easily find it, but answer-writers have more difficulty - until it's provided. That is, to my understanding, the primary reason for questions to have a "closed" status. (I would like for default views to filter out such questions, while making it easy to include or even focus on them.)
We can adjust policy by site volume
As it stands, we get so few questions, primarily from a community of relatively skilled enthusiasts, that anything that doesn't measure up to standards sticks out like a sore thumb (and there are plenty of resources to deal with the problem).
If Codidact communities are to grow in the way that we seem to hope they will, we must accept that this will not always be the case. Any discussion like what we're having here is inherently forward-looking. I can see a path that leads all the way to "new questions start out closed or in a separate section, and must be workshopped, categorized (as potential canonical material vs. yet another 'support ticket') and approved before they can be published to receive answers".