I'm a mod at Literature, and one of the things we struggled with is that early on in the site's history, people had a very narrow conception of what types of questions could successfully be asked and answered. People were writing all sorts of posts on meta, and writing all sorts of messages in chat, arguing about what sorts of questions should be on-topic or off-topic. And people, on the basis of no evidence whatsoever, came to the conclusion that various sorts of questions should be off-topic.
Then of course, people asked those "off-topic" questions (remember: no one reads meta), and it turned out that they could be answered successfully and that they had a place on the site.
Questions are closed because they can not or should not be answered. The interface for closing questions gives five reasons for closing questions: a question is a duplicate, a question is opinion based, a question is off-topic, a question is unclear, and a question is too broad.
These close reasons were not chosen arbitrarily. Rather, the history of Stack Exchange is in part a history of learning what questions work, what questions do not, and modifying the close reasons accordingly. These close reasons were chosen because the Stack Exchange community learned, through experience, that these questions would not work.
You could have some conversations on meta and arbitrarily, on the basis of no evidence whatsoever, determine that entire classes of questions should be off-topic. Of course, only a small percentage of people who use the main site read meta, so you'll be excluding a huge portion of the membership from the decision process. And you'll run the risk that maybe some of the questions you exclude are actually great questions.
The other option is to actually ask the questions and see how they work in practice. If they don't work, well, you have close votes and should know how to use them. And when you do make a policy about closing questions, you'll have the advantage of knowing that it's the right policy, because you have data to back it up. And if they do work, then congratulations, you added another great question to the site. And because closing questions is a main site feature, you won't be excluding everyone who doesn't visit meta from the decision process.
I personally prefer the second process. But eh, it doesn't really matter in the end. If you go with the first process and write a bunch of meta posts, those questions will still be asked again, because no one reads meta. But then you'll have to go through the process of closing the question, and maybe some people, on seeing an example, will change their mind, and other people who haven't read the meta post might disagree with said meta post. So then it will turn out that the meta post is useless and doesn't actually reflect what happens on the site, and you'll have to write a new meta post to reflect the new consensus, and you'll find yourself wondering why you went through all the trouble of writing the old meta post that wasn't even based on any sort of evidence.
The moral here is: don't waste your time writing meta posts about hypotheticals. Have actual data first.