I think it's a holdover from hiring more junior candidates: after a few years interviewing and sitting on HCs at Google, I hired for a tiny startup that had tons of trouble sourcing candidates. My interviews at the latter place eventually devolved into basic IQ tests. Simple coding problems are pretty ideal for this, since they also lightly test necessary domain skills.
The problem is that this doesn't scale with talent and/or experience: I certainly wouldn't be better at coding interviews than I was out of college. People who aren't great at hiring take an approach that's valid at the low level and extrapolate it well beyond the bounds where it makes sense.
On the other hand, this only works if the pedigree is recognizable and reliable. I can be pretty confident that I don't need to test basic coding literacy of someone who was an engineer at Google for four years, but I've interviewed people with ten years experience and great references who were clearly dumb as a brick during the interview. In one case, we hired the guy over my objection and spent the next year fighting the fires he'd set whenever he touched anything (the founder eventually found some relatively unimportant, isolated niche for him to keep busy in).
It's a really hard problem, and most complaints I've come across are generalizing from their skewed view of what the candidate pool is like (usually assuming that most candidates are roughly as talented as they are, when the reality is that many candidate pools consist of some of the dumbest people you're likely to ever encounter)
[1] ie, practically everyone