White would probably have preferred 22.h4, but for 22...Bxh4. 22.f4 seems to be safely met by 22...Be4 White's sole chance, however slender, lies in a pawn storm on the K-side. That already seems to be stymied, pretty much.

I wonder if Be3 instead of Rxe8 would have worked?

H'mmm ... I would not have played this! Now with 28.f5 White could have presented Black with a very nasty technical challenge. When defending, "bad" bishops often come into their own, but Black also has the problem of finding a route for his king into White's position. Nor has the light-square Bishop any means of ingress. After 28.f5 White can think about sealing off the Q-wing with b4 and (maybe) a3. After 28.f5 I really fancy White's drawing chances!

In general, White would certainly be intersted in exchanging pawns, but not here. There are just too many pawns. As Thomas observes (and see my last comment) 28.f5 was the move to make Black's winning chances more problematical.

The exchange does more than free the dark sqare bishop. It removes the short term problems of the light square bishop. Black doesn't have to work nearly so hard to activate the light-squared bishop.

My first thought was that this move was absolutely wrong - but then, what do you see after 29...hxg5? How is Black to break into White's position with his King? With 30.b4 followed by a3, White leaves just the h-pawn as a target for the Bishop. This he can defend by oscillating his K from h2 to g3. Black's sole chance lies in his a- and c-pawns. If he can break up White's Q-side pawns, which will involve a sacrifice on c5 at least, then he can break through. Black still retains the edge here, but it isn't going to be easy!

White has two possible plans, play for the win or play for the draw.

30.b4!