Then it's a test, not an interview. I don't think interviews were ever supposed to be perfectly objective. If the interview is only based on CS textbook questions, they might as well replace the interviewer with an online test, which by the way increasingly happens for screening, but my point is that it filters out great engineers who worked and delivered solid products because the interviewers are just not interested in details of past work experience
Again, yes. It is hard to fire a bad candidate in case the interview was misleading. Companies would prefer false negatives over false positives. What you see in interviews is a product of laws about hiring and firing.
If this is the case (i.e. companies can afford this behavior), all the claims about skill shortage are a lie and should be called out.
in-depth questions
How is that any better than hiring a terrible dev that just spent 3 months memorizing Leetcode solutions? Aren't they essentially being dishonest about their skills?
>time-consuming Phone screens are already several hours over several days and on-sites are typically 8 hour affairs. Is this not already time consuming?
Genuinely confused by the Stockholm Syndrome-like mindset that devs have for Whiteboard interviews.
memorizing Leetcode solutions
I have my doubts. Indeed from my own experience with Leetcode is I can only do well on the questions that I think quite hard about. Recently I tried to do a Leetcode question marked Medium, but my 3 line solution (while correct, and working on my own test cases) always resulted in "Time Limit Exceeded". I studied their test cases, and realised that my brute-force solution was of exponential time complexity. I then read of the discussion of the problem, was so delighted with, and surprised by one of the proposed solutions (linear time) that I felt compelled to write down the loop invariant in Hoare logic! Fun evening, and I'll never forget this neat algorithmic trick. I think that practising enough Leetcode so as to do well on the Medium/Hard questions requires developing an understanding of algorithms which is highly useful in practise, in particular, if you work with large data sets, clever algorithms are unavoidable for getting anything done. Have a look at the list of past ACM coding contest winners, e.g. the 1993 winners [1] include Craig Silverstein (first hire at Google) and Tony Hsieh (Zappos, LinkEchange).
Summary: I don't believe that doing 3 months coding challenges is meaningfully described as memorisation.
On top of that, doing coding challenges is perfectly compatible with asking other questions.