Wrong Puzzle: 1209936

Today in solving puzzles I faced this puzzle:

https://www.chess.com/puzzles/problem/1209936

I explain it why, it might be interesting lesson for those who are not good at endgame understanding:

This puzzle is wrong. After 1. Kg2, Rb87,2.RxN, Qe7, instead of 3. Rc7, black can play 3. Rxf1. The point is that after 3. ..., QxR, 4. Bd5, RxQ, 5. BxR, Qxb3, 6. axb3 then black either has to move something in kingside or the "a" pawn. If it moves the "a" pawn then it is easy to see that white's king reaches the queenside quickly and wins the pawn eventually and creates a passed pawn. The only option is to move something in the kingside and in particular the king. In this case white first plays a prophylactic move b4, to stop black's a pawn from advancement, then white and black have to fight for kingside advantage, but since the pawns are in their early stage none-of they succeed to achieve any advantage in the kingside, however, white manages to fix the kingside pawn structure. Then white's king starts to march to the queenside, the black king also has to do the same, here the problem for black is that 1. white has a space advantage (because of b4 pawn the blocks the black king to come from c5 or the "a" pawn to move to a5), 2. white has an important tempo b3! while black doesn't have any tempo without fixing its own pawn structure. White easily wins this endgame (even if it makes small mistakes), almost every road is in favor of white.

Seeing all this as a human is not difficult: I'll fix the pawn structure in the kingside, then I have space advantage in the queenside as well as one important tempo. If kings are in opposition state, as a white I can simply win. But engine first only realizes +1.5 for white, then after pawn structure is almost fixed, goes to +3-4 for white, then once the kings are marching towards queenside, it realizes that white wins. So this tactic has multiple solutions, and either should accept all of them or should be removed.

While that may still be winning for white, the solution has white up a piece and has a much better overall evaluation.

@Martin_Stahl

A puzzle should have exactly one solution or in other words, it should stop if there are several winning branches. All I explained was to show that the computer couldn't see immediately that my solution is certainly winning. It takes time for the engine and just after you make enough many moves, it realizes that yes my solution is absolutely winning. So the current puzzle is actually wrong. There are two solutions for it, one easy by human eyes (at least by my eyes) but hard by computer eyes, the other easy by computer eyes and maybe not so easy by human eyes. None of them has an advantage over the others. If my solution was bogus, then it was another story, but even engine after a few moves agrees it is a clear win.

Relying on the engine's initial evaluation is a mistake, there are positions that engines don't see immediately but human player finds the logic behind it easily.

You have a point there and actually the problem is worse. Considering that the puzzle challenges are basically tactical, it is fair to expect that some combination will take place. Rxf7 is precisely such an event which wins on top of that in a fairly easy manner - by human standards. A higher evaluation score does not always concur with an easier win. So it is arguably a bad puzzle as you suggested. The engine is excused by the circumstance it is unable to mimic human logic, as are we for not appreciating engine evaluation scores. One conclusion could be that no puzzle should be entered into the database without prior "eye" inspection on at least IM level.

Another Stunning Puzzle, imagine that I could not solve K+R vs King and this puzzle is rated 2200+:

https://www.chess.com/puzzles/problem/1376154

What's this? Is there anybody to make the least sanity check?

Anytime you see a puzzle that is basically a mate, no matter what, it's looking for the quickest mate. There are actually a few of those and if you look at the FEN you'll see that if mate isn't completed in X moves (most of them I seen have been 5) then it's a draw 

edit' this one is 4 moves and anything longer is a draw.

@Martin_Stahl,

Why anything more than 4 is a draw? Mate is mate, 4 moves or 10 moves doesn't matter. Rook and a king is mate, at least if they want to get exactly in 4 moves, they should say mate in 4. Are you serious about defending a nonsense puzzle like this? Every puzzle should have a unique solution, period. There is no other option. Sometimes some puzzles like the first one have 2 winning possibilities, they should stop it before any branch.

But R+K against a naked king has millions of possibilities, even the first move can be almost anything, then I'm wondering how you defend this instead of reporting to delete it and refunding our points?

50 move rule