This week I reviewed Aronian-Caruana Sinquefield Cup 2017, in which Bishop + wrong rook pawn was a possibility. I watched this game live on site in the no-engine GM discussion room. During moves 45-55 (~30-60 min total), GM opinions about the position on the board went back and forth between Draw and Win for Black . (GMs were V, Akobian and E. Hansen). At the time I vaguely recalled that Bishop + wrong rook pawn is surprisingly a win if the opponent has doubled Knight pawns. Based on this I claimed Win for Black somewhere around move 52. This evaluation was subsequently and indirectly confirmed by both participants when Fabi offered the R exchange (move 56) which Levon declined.
https://www.chessgames.com/perl/chessgame?gid=1883502
===My vague recollection has now been traced. It’s from a book I’d studied a long while back, Batsford Chess Endings, page 121, which analyzes the conclusion of Short-Kasparov Belgrade 1989 starting from move 93. Kasparov demonstrates the winning method.
https://www.chessgames.com/perl/chessgame?gid=1070432