Your reasoning for why Leetcode is bad is a bit flawed. The reason why it works is because it's correlated with other performance. The fact that a customer don't ask for it before they give you a job doesn't disprove the correlative factor. The fact that people snub it by thinking it's too far removed from reality doesn't make it so either. The fact is that someone who is clever enough to consistently do things like reverse a linked list in one line quickly and correctly will also be able to quickly do less clever things correctly.

Your idea is that only doing the exact thing shows if you are good at it. You are almost implying there are no correlations between skills. If that's what you believe then looking at the research on IQ should clear up that empirically that is not the case. And note I said the research. Not news articles or editorials or blogs or podcasts. The actual scientific literature.