Thhere are always some counter examples, no interview process is perfect, but the reasons people might not do well in their job are extremely varied and a lot of it is simply the work environment . Don't let personal anecdotes create a false sense of data. The fact is, statistically these kinds of interview questions are extremely powerful predictors performance.
The pure IQ tests and things yes, but the coding tests are never done in your own time. that is the whole point because otherwise you would just copy and paste an answer, so again it sounds like you don't really know what the goal of these kinds of tests are.
ehh, these tests are done face to face, albeit usually virtually now unless you are hiring close to an office.
You are also completely failing to acknowledge that problems like leetcode tests are only part of an interview process, as was pointed out to you several times. Having a systems architecture interview could be part of the process, but A) such an interview is only relevant to more senior roles, and only in some domains, B) There are plenty of people that can talk the talk but can't walk the walk so any interview based upon purely describing some system is a weak predictor that has risks. Nothing beats getting someone to actually code , you can split the wheat form the chaff and find out those that cannot only communicate well but can provably put pen to paper and show results.