I have these I have used "challenging" algorithmic stuff a handful of times in my careerer. Same with recursion, I have used it, but not too many times.

If I do hackerrank or leetcode or something similar these days, it's pretty much luck if I will get it. If I sat down long enough I could work it out, but with a 2 hour time limit, its usually 50/50 whether the answer appears in my head in that time frame.

And here's the thing. I write a lot better code than I used to. I know what is important, and it isn't clever algorithms, it's choosing the right abstraction, and keeping things maintainable. Making the code easy for the next developer to follow. For the vast majority of software roles out there this will be so much more valuable than algorithmic knowledge. Sure knowledge of computer science fundamentals is important, but I think we give it too much emphasis over what really counts.

Basically these are testing for a computer scientist when the job is for a software engineer.

It would be interesting to look at the correlations between the two, as I'm not sure if that's every been done.

At some point there's a signal to noise ratio issue here as well.

If companies are passing on specific pockets, it’s likely from a belief that the conversion from top of funnel to successful hire is low.

There’s plenty of excuses for gatekeeping, but the real reason is laziness. Of course, we don’t call it that, we call it a meritocracy and apply our biases as “common knowledge”, “proven”, or if we’re feeling particularly full of ourselves, a “sufficient statistic”, but never what it is: a bias, a blanket assumption, a gate.

As soon as you’re relying on assumptions, it’s not a meritocracy. It’s bias promotion, ie gatekeeping.