Lately, I'm seeing a lot of questions in the review queue which are marked for closure due to being 'too localized', or 'not a real question'.

Some of these are marked on the same day they're asked, and some already have answers.

If the question is bad -- fix it and/or comment and/or or vote it down.

(As Luis Alvarado suggests, showing a new user how the site should be used, how good questions are written, or giving other constructive guidance in comments is ideal.)

If someone is quickly able to answer a question, it seems likely* that there was a real, answerable question. If it has generated a quick response (upvotes, comments, good-bad-mediocre answers), it seems likely* that it affects more than a extremely narrow subpopulation, and it certainly hasn't been abandoned. (*I'm assuming answerers are rational, altruistic, and that other optimal (if unlikely) conditions are being met -- but the general point remains.)

Some of these questions are being flagged so rapidly that there is no way that a reasonably large group of potential editors or answerers has even seen them.

Even if a question is astoundingly ignorant, it still may prove enlightening for the similarly placed searcher.

If a question is badly expressed but still obvious, or layered under meandering narrative, or not explicit, just fix it: edit it, leave a useful comment. In some cases, vote it down. On seeing a really bad question, you can even just run away. All better than voting to close with slim justification.

Don't use the option to close as a cudgel, a reaction to a substandard or badly formulated query. This prevents any possible good coming out of bad but relevant questions. (And if it's a bad question because it's a duplicate, take the extra 75 seconds to find the duplicate and mark it as such.)

I don't know if there are good counter-arguments, but I think this practice is harmful and misuses the tools we've got.