sachistu: Even though ChessBase and some other sources give this version of the score, it is apparently incorrect. According to a diagram in the Hungarian magazine Magyar Sakkelet (1989, p10) which gives the position after Black's 34th, the White pawn is on f3, not h3. Working backward suggests that White's 22nd was f3 not h3. Using the rest of the moves as given does result in a mate after 41.Bh6+ (as also given in Magyar Sakkelet).
A second point in favor of 22.f3 being the move is that in the score given here the e-pawn is hanging on move 29. That possibility does not exist in the version I am suggesting is the correct one.
In George Rajna's notes to the game in Magyar Sakkelet, he indicates move 35 (c5) is dubious, and suggests 35.h3! Of course, this would not be possible had White played 22.h3. Rajna was in the tournament, so it seems reasonable to suppose he had access to the score(s).
Rajna gives the following line as winning: 35.h3! Nf6 36.c5 bxc5 37.bxc5 Ne8 38.cxd6 Nxd6 39.Qc3 winning. However, I think Rajna missed a better defense with 37...Qd8!?.
At the end of the game, Rajna points out Black was in time trouble (accounting for the blunder 40...Kg7??) and suggested 40...Ne8 instead.
Hopefully, this clears up the questions raised back in 2011.
I'll submit a correction slip unless there is some objection.