I rather like White's move in principle, but maybe it needs to be prepared, e.g. by a preliminary Bf2. Not an easy decision, though, as the bishop seems to be carrying out a useful function where it stands, covering the d8-square.

33...Rf8!! cuts the White king out of the game. White could try to unpin his bishop by 34.Rd2 but 34...Rf4 forces off a pair of rooks, whence Black's remaining rook clears out the Q-side side: 35.Be3 (say) Rxd2 36.Bxd2 Rxa4. 34.g3 stops the rook's invasion at f4, but the bishop will soon fall.

Perhaps White could have tested Black's alertness by 36.e7, but after 36...Rd7 followed by 37...Rfd5 it's all up for the bishop. Having already seen the annotation by qamar1, it is interesting to get the perspective, as it were, from "the other side of the hill".