Sadly, yes, at least for a certain type of job. Probably not the best, but considering the effort involved on the part of the employer (ie very little) combined with the result, it's pretty good. Anything better requires a ton more effort (work trials) and/or doesn't scale well (personal referrals).

If you consider that >10 years ago there were blog posts talking about how hard the "FizzBuzz" question was to implement when asked in-person as a way to demonstrate how you approach solving problems, there's definitely a reasonable assertion that these types of interviewing methods help weed out applicants unable to write basic compound statements. Might not be the best way as a single pass/fail criterion though.

Sure, if your company is as prestigious as Stanford or MIT.

Leetcode style interviews were always a reasonable option for a particular type of company. But they became a trend for companies that weren't a good fit.