The majority of problems I've been working on in prep fall into 1) do you know this algo, 2) can you manage n-pointers in an array/set of arrays, 3) can you keep track of multiple deep stacks in your head (deeply nested/multiple loops, recursion), 4) can you come up with clever micro-optimizations?
As far as adaptability goes, I also didn't see it. Adaptability is better measured in a more real world task like subtly expanding a requirement in a series of logical steps.
You can understand a candidate's use of data structures by asking them to build a TODO list API. To test for adaptability, now add tags. Add sorting. Add sorting by various fields. What would you index on if stored in a DB? How would you make this service scale if servicing thousands of users? Millions? Make it a shared TODO list with multiple users.
Want to test full stack? Add a React UI to this API.
Look at that: see how a practical coding interview without leetcode that can also be extended to test systems design? A simple TODO API and we can keep extending it to more deeply understand grasp of data structures + real world adaptability to shifting requirements and how to code for it.
Leetcode's problem is that it is so far removed from real day-to-day programming that it has to be studied for explicitly. And therefore encodes it's own bias. Leetcode also misses a lot of key measurements of experience; I've yet to see leetcode exercises on proper exception/error handling. Never seen a leetcode problem on how to write more modular code. Or even really practical secure coding exercises like testing for valid inputs and protecting against XSS.