BUG - Zero Inaccuracies evaluation, despite blunder

@Toadofsky Thank you very much for sharing. i will consult this wealth of goodies.
I will also digest what you wrote in your post as i go through those references. thks.

Let's take a step back to revisit this from an overview.

When I click Request Computer Analysis, I'm doing so to see all the times I made a blunder, a mistake, or an inaccuracy. That is objectively speaking what the feature is for. Subjectively, I interpret this to highlight all the times I made an incorrect or interior move, when a better move was available.

We can further simplify this discussion into asking a simple yes/no -- should 24...Re2 be acknowledged in the Inaccuracies/Mistakes/Blunders tally as an error. I say strongly "yes", it should be counted, but some have said "no" it shouldn't since black is still winning.

The current behavior is completely inconsistent with my intention for requesting computer analysis in the first place. I am using the feature specifically for this type of a scenario -- to alert me that there was a far better move available. I want to know if there is a better move available, regardless of whether or not I'm still winning.

Everything I just described is simply stating the obvious and is not controversial.

If you actually analyze the position after best play by both sides 25. Rxf2 Rxf2 26. Be3 Re2 27. Bg1 Bxf4 28. c4 Nd6 29. cxb5 axb5 30. Nh5 Be5 31. Ng3 Rxb2 32. Re1 Re8 33. Rd1 Nc4 34. Ne4 d6 35. Nf2 Rg8 36. Rd3 Rf8 you'll find Black's attack completely crushing. Therefore 24... Re2 is one of the best moves.

@Talmanian To put it bluntly: the system is not tailor-made for terrible play. It works fine for everyone except people who play with a +/-15 position for 20 moves. Objectively no real analyses is required, just try and take pieces when your opponent gives them away for free and try and put your pieces next to the opponents king, perhaps you can even win by checkmate!

@Talmanian regarding #22: as a lichess user, my desired behavior when I click "Request Computer Analysis" is different than yours. I recall the analysis previously did what you described: it produced long lists of "mistakes" during completely won endgames that were a complete waste of time. My analogy: if I want to improve my golf game, how much time should I spend practicing my 6-inch putts, to make sure they are always aimed at the exact center of the hole?

Furthermore, from the perspective of trying to win a chess game, the thing that matters is capturing the king. In a line that leads to winning the king, sacrificing material is irrelevant. Experience indicates humans are prone to over-value material (such as your queen), but a bug-free eval system should be aimed at predicting ability to win the opponents king. From that perspective, it would be a bug to flag positions based on immaterial pieces on the board.