Look at us: two senior technical leaders with deep software backgrounds arguing about minutia - we're our own stereotypes!
I pushed back on the confirmation bias assertion because subjective ≠ biased by definition. That's attribution error, plain and simple. When something feels different, it's an invitation to try to understand what's happening. Hence my OP. Bias is just one of many possible explanations.
As to chess.com's developers - I see them advertising C++ devs, backend (Java/PHP) devs, app devs. I don't see a lot of AI action going on there... Maybe there's some very specialised NNUE stuff in their engine teams (e.g. Torch). You can get a long way with a stockfish binary, some opening books, some fairly dumb heuristics, and a PRNG. And that is how the chess.com bots feel to play against (yes, just another hunch) - or at least the lower Elo ones.
Took me several attempts to beat potato salad.