I've spent my entire career solving real world problems, but it always comes back to tricky logic puzzles or deriving best case runtime complexity on paper with these interviews even still, and even for non-FAANG companies. I regret not just reading cracking the coding interview every couple of years and passively leetcoding at this point because I still might fail this technical screen. And that does suck.
Can you follow a recipe? If you're missing an ingredient, can you employ a substitution? That's what an awful lot of programming is like, except using keyboards instead of stoves. Now, it still takes some talent - a talent for computer "recipes" rather than literal cooking recipes - but it's not as out of reach as leetcode leads you to believe.
Do you have evidence for this claim? Specifically, that the skills and talents that make one good at Leetcode are not correlated and predictive for being a good programmer.
I actually never heard of Leetcode before this thread but I highly doubt that. This seems to be a puzzle site like Codewars or Codingame. While puzzles like that are fun, they are only one small part of being a developer. Design, architecture, UI/UX and communication are a lot more valuable than being able to write your own Sudoku Solver.
Unless you work in some cutting edge project, optimization or framework development, most of your work will consist in performing api calls, copy pasting and glueing boring pieces of code together.
So, my recommendation is, instead of becoming an expert in solving code challenges, be an expert in implementing real world solutions. Learn some useful framework and try to build something useful with it.