In the game Isaak Boleslavsky versus Vassily Smyslov of the USSR Championship 1961 the following same colour Bishop endgame was reached. On move 46, the position is WHITE King on b3 Bishop on e3 pawns on b5 d5 e4 f2 g2 h3 BLACK King on e7 Bishop on d8 pawns on d6 e5 f6 g7 h7. BLACK to move.
[FEN "3b4/4k1pp/3p1p2/1P1Pp3/4P3/1K2B2P/5PP1/8 b - - 0 1"]
Soviet commentators hailed this endgame as a Smyslov defensive masterpiece (the game ended with a draw) and Aveskulov agreed. But Stockfish 14.1 NNUE depth =66 seem to indicate White missed a win. Indeed, after several minutes of analysis Stockfish changed its apparently fixed evaluation of +3.60 and evaluation starts to go up. Black than switched to new system of defense, but the evaluation remained then at +4.14. I tried then to follow some moves in the line indicate by Stockfish and did a second iteration some moves forward. It seemed a White win indeed (evaluation at +6.93 at comparable depth).
What happens with perfect play? And why did no commentator pointed to at least a potentially approach for White? In the end, Soviet grandmasters considered this position a fortress, But Stockfish 14.1 NNUE disagrees. Why?