Comment by hilux

Scenario (1) is trivial to detect.

But Scenario (2) is rather complex - most strong players who are "smart cheaters" would use the Stockfish suggestions infrequently, perhaps just once per game. The difference to the game result can be significant, but detecting this sort of cheating is a statistical exercise.

As you said, it comes down to thresholds, and setting a tradeoff between false positives and false negatives.

And let's not forget that a false negative, which is the default case and extremely common, also has an effect – it disadvantages all the honest players who lose to a cheater.