I m not sure that exercice 61830

All is in the title :
The finale position is Q vs R & 2B but the all is closed for the white color bishop of black and the queen paralyse the black position.
After h4, g4, g5 i thing is probably draw.
Of course black is certainly better and black game is hard but the win (if exist) is not so easy
fr.lichess.org/training/61830

After final blacks move Rxc2 white have to move Bc1 or Rb1 and in both cases then he have his Rook blocked defending the bishop.
Now after Bc1 or Rb1 black move g5 to avoid h4 or g5.
Then, in the future, black can push also with f5 in order to move his e4 pawn and sblocking his b7 bishop.
But you are the top 200 player, not me.
So i'm asking to you if i'm just saying a lot of bullshits xD

I had seen that puzzle before. It's a nasty one, and underrated in my opinion. I failed it because it made no sense to me at first sight. I don't think it's wrong, but it's certainly hard and weird. I'm not sure any GM would go for that kind of endgame in a real game. Besides, it doesn't feel like a puzzle since you just give up your queen to play a position that's far from clear and without any immediate blow. Of course after you play some moves it starts to make some sense but understanding it from the starting position is just too tough.

@sheik_yerbouti
after Rxc2, Rd8+ (win the Queen)
Computer say that Bf8 is then better that Qxd8 and i agree but the position after that is not so clear.

if you saw what i wrote forget it

Lol this is damn true! I did not saw this move xD
Now the exercise seems to be totally nonsense, it makes you pray of a terrible threat that you can surely evitate doing Rxd1 instead of Rxc2 as final move.
Anyway if you see the game from which the exercise is taken, you can see that final moves are different it.lichess.org/dsLaMOrA/black#39
Maybe exercises are created by a computer algorithm that finds crucial moves of a game and try to cleans it from mistakes in order to make the situation playable, but mainteining the aim of the crucial moves (in this case: take a bishop) and sometimes this algorithm paradoxically leads you towards wrong endings because of you achieve your goal and the puzzle ends, but if you had continued to play you would have abandoned.
Ultimately exercises are not perfect yet, but in most cases are helpful and for this reason are a good tool to get better.
If you want you can find a lichess moderator and ask him about how exercises are made, but surely this one is wrong.

In this position there are clearly more better moves than the final train. It is a pity that these alternatives are ignored and the task is not solved correctly.

HOW DARE YOU QUESTION THE MIGHTY STOCKFISH??? I bet it knows lil bettah

All puzzles are designed to be solved the way stockfish would play them at a certain depth level. for ex, in result you get +6.9, but your way of solving ends up with +6.2 or +4.7 which, from human perspective, by all means is significant advantage, but it seems like stock-badass-fish doesn't give a damn you know what. VARIATIONS mode where your point is to get an advantage out of a particular position is not available yet. Would be a pleasure if such a beautiful option comes out one sunny day.