sachistu: In this sharp game, an interesting position could have occurred after 16.d6!
Instead of the game continuation (16...exf4), one of the alternatives analyzed was the following: 16...e6 17.fxe5 Nxe5 18.Rf4, when one of the possible continuations is 18...Nd5. After 19.Nxd5 exd5 20.Nxf7 Nxf7 21.Rxf7, Ufimtsev and Furman in the tournament bulletin gave 21...Kxf7 22.Qxd5+ Be6 23.Rxb7+ Kg8 24.Bc4 leading to mate.
However, in SvSSSR, 1954, an alert reader pointed out that after 24...Bd4!, it is Black who is winning! Actually, Ufimtsev and Furman were right about White winning as instead of 23.Rxb7+, White has Rf1+! and after ...Kg8 24.Rf8+ Rf8 25.Qxe6+ Rf7 26.Bc4 does win. Unfortunately, their analysis turned out to be faulty.
Apparently, Terpugov realized that taking the Rook would be too dangerous for Black as he gives instead 21...Be6! in the tournament book page 78-80.
Geller was rather fortunate in this game as Terpugov was winning until two consecutive blunders allowed Geller to escape. On move 24, instead of 24.Rf3?, 24.N3d5 is winning. Likewise, the same move was still available on move 25 (instead of the blunder 25.Qxa6??)