white 29.Bg5 seems contrast with the natural moves.
It seems to trap his own bishop, and to be honest the white position is worse (slightly) than black. But actually 29.Bg5 is a trick to lure the black g-pawn and (hopefully) f-pawn to advance and leaving the e-filed bishop unprotected.

Amazingly Kramnik thought this in the blitz game.
Assuming
29.Bg5 f6
30.Bf4 g5
31.Bc7...until now the position seems to be in the blacks favor although slightly, and if 31...Bd7 (hoping to trap the bishop). then

32.e4! Rxc7
33.exd5 Bf7 (Bd7 bad, because after white queen retreat ((e.g 34.Qc2 Qc8 35.d6, leaving both black rooks hanging)) 34.d6...now the position seems to distort the table (although slightly) with both of black rooks hanging.

I dont say that 29...f6 is bad move, and in the game black 30...Rac8 also nothing near the badness itself. its impossible to accurately analysis the variations.

Actually, for all the sum of the moves in this game lasted, black was in the better position for the 80% of the moves.

You need only one bad move to ruin your game and regret it for the rest of the day, in this case namely 87...Rxb5+??...its such a careless move, white can achieve draw even win with other secure moves (Ra2, Ra1, Ra8).