My thoughts on the question
The only reason I can see for an existing answer to affect a question's MSE on-topic nature with respect to the guidelines is that it actively refutes a particular close reason. For example, an answer citing facts and studies works against a question subsequently closed as opinion-based, and an answer that uses an overarching concept or a universal trick to tackle two or three apparently separate questions in one fell swoop works against the argument that the question is broad.
All votes do is attract my attention to some answers and make me analyse, more carefully, whether they refute the applied close reason. So votes matter, but more importantly the nature of the answer as compared with the applied close reason matters.
Closure does enough to perhaps redirect people elsewhere and send the need-for-improvement message. Deletion, on the other hand, is a measure that removes questions which haven't been active for a long time (via the community user) and nullifies the visibility of a closed question, preventing it from receiving any attention from the majority of the community and leading to loss of reputation, especially if it's a recently created post.
Mere closure doesn't seem to have been enough to deter the answering of low-quality questions. Deletion, a stronger measure, has a better effect in curbing this habit, per empirical observation, because of reputation loss and the accumulation of such posts leading to moderator action ref. EoQS. Deletion is therefore a deterrent to the answering of low-quality questions.
It is clear that I placed an application of the guidelines over the well-received nature of the post and the efforts of all participants on that thread. So why did I do that?
Write a new question?
The first is that I am not sure why people are not using the best method to rescue content : a new question. The insistence of the methods of improvement-prompting and context-rewrites are historically unsuccessful, the former because of the asker being a participant and possibly not having enough interest, and the latter because of the balance between "words in asker's mouth" and "too little context" being very fine and open to interpretation as I see it. These methods should only be used if one is very patient or very confident.
The only problem I see in asking a new question is that the question could already be off-topic for another reason (which would likely be the case for the earlier question as well) or the original question could have more than two answers , which are difficult to copy across to a new thread. Given that most of the answered deleted questions I observed had precisely one answer, this captures the majority of such questions. Furthermore, such a process of rescue scales in numbers, as I explain below.
There were roughly ~$400$ questions asked yesterday. Of these, about $85$ were closed, giving a rate of roughly $20$%, say. The deletion rate is slightly lower, around $19$%. Of a randomly chosen list of $30$ deleted questions (note : continuously chosen from a list of questions deleted about six hours ago) , $23$ were unanswered, $7$ were answered with $5$ having single answers (of which all had vote either $0$ or $-1$) and one had two answers, each with a $-1$ and the last one was actually well-received with $2$ answers, one with a $+2$. Most were community-user deleted. When scaled, there are likely four questions in a day which are deleted with an upvoted answer : easy enough for individual-question rescue. Finding them is perhaps a bit more difficult, but the daily number is quite small.
Part of the reason I favour deletion is because I want to encourage people to rescue content from these questions by writing their own questions with context added, thereby raising the question's standard and allowing the rescued answer to fit more hand-in-glove with its rewritten question. I would like participants and viewers to redirect their efforts to this form of rescue : generally uncontroversial and successful.
On votes
The other has to do with what I think of votes. What exactly does one vote for? I think one should base their votes on how much hand-in-glove the answer is with the entirety of the question.
For example, let's say there's an answer that refers heavily to the user's context but is somewhat long and labouring, while the other is a slick one-line answer but doesn't comment on why the OP's effort is misdirected or not enough. Which would you vote up? Perhaps both : but I would definitely vote up the answer that appreciated the context provided by the user and commented on it in their answer.
Context is, after all, supposed to help someone address the question to the best of their ability, so to see someone not so much as go near it but get $10$ upvotes, while the other one gets just $2$? It's natural because we gravitate towards beautiful and slick math, but it isn't appropriate for MSE if we are choosing to have people add context. Which sounds weird, but true.
If votes were directed at answers that worked the hardest at harnessing the user's context and attempted to nullify any potential reasons to close the question, then we'd have an immediate reason to see why well-voted answers can easily deter deletion and closure. For example, if a question is broad, an idea that somehow brings the entire question under one umbrella should get credit over three distinct tricks. For questions which may be opinion based, you would vote for the answer that has enough facts and citations, and not for the post that you relate with.
Therefore I prefer deletion over preservation of highly-voted answers : because the way people seem to vote is natural, but often doesn't favour and promote answers that actively work to use tools from the question and work on the potential site-based pitfalls of the question. That means I'm in a minority, but I can see how well this notion of voting yields a more positive enforcement of the guidelines, and would therefore defend it all day and night.
