Interesting is that possibility to play Rxd4 idea for putting rook on c-file is going to be on board for quite many moves from here on. It seems like white can hold the draw out of it.

Yes, this is inacuurate because of 23...f5

Watching this unfold, one senses uneasily that White's game is gradually becoming more and more constrained. Already Black's mobility is the greater, and after this White begins to look constricted...

Well, it comes on consideration Bh3, pinning f5, and with the idea of g4, just to open the game and providing squares for the rook on the 3rd rank.

Another - albeit odd-looking - idea was 25.b4!? If 25...Bxb4 26.Rxd4 leaves the pawn position symmetrical; and 25...Rb8 26.a3 places some constraint upon Black's Q-side: 26...a5!? 27.bxa5 Rb5 28.a6 Ra5 29.Rb1 Rxa6 30.Rb7ch Kh6 31.Rd7 and this rook's activity indirectly protects the pawn at a3.

All is not lost though...

This is OK, and goes with a defensive plan that is very hard to contemplate for the inexperienced, though I'll bet battlefleet had his eagle eye out for the possibility: 27.Rxd4!!? It is possible that 27.b4 is better, but I do like this exchange sac, breaking out of the chains: 27.Rxd4 cxd4 28.Rxd4 g5 29.Rc4! Rc8. After this, White has if anything the more active game, and it will be very hard for Black to find any kind of lodgement in White's position.

27.Rxd4! seems to be enough. That possibility has been on board for sometime and the same idea should work even if played earlier for example 22.Rxd4.

I would almost certainly have played 28.Rxd4 in this podition, but 28.b4 was also worth considering. Sometimes passive defence is the best course. I don't think it is here. Mind you, it isn't always easy to tell!

the GK computer is in error.

not necessarily. the gk computer takes pure brutal logic into account and not players talent levels. the rook on b8 prevents further advancement of the pawn but not much else.

What does the maching suggest as better: 28...c5!? Can't say I like that idea at all!

Well, it might be that computer wants to play 28...c5 to stop any Rxd4 plans(?).

against a master giving him a rook advantage probably would have been very bad indeed. Still there would have been options...

No: I reckon the machine has the right of it. In fact that exchange sac - as I have already remarked - has been in the air for quite a while, now.

Although it's probably OK - maybe - I don't like this move. Since White has adopted a passive defensive stance (which, contrary to my earlier expectation, is working out pretty well), then maybe he ought to close the position right up by 31.h5. As Black is hardly likely to take on h5, nor yet, by allowing an exchange on g6, to saddle himself with an isolated pawn upon a file already occupied by an enemy rook, his likely response would be 31...g4. After that, the only way that Black would be able to lever open the position is by [a] bringing the King over to a3, or [b] a bayonet charge with the c-pawn. Neither plan looks very promising.

Possibly your instincts were telling you something. Clearly, ...c4 is now high on the agenda. Moving the rook one extra square over (Rb1) anticipates this advance, and says to Black 'Come closer: I dare Ya!' Why? 38.Rb1 c4? 39.dxc4 Rxc4 40.Rb6ch Kf7 41.Rb7ch Kf8 (not 41...Re7?? 42.Bd5ch) 42.Rb6, attacking the K-side pawns.

Had the rook stayed on d3 and the other rook moved to c1, this resource would have been unavailable to Black. Not that it changes all that much after 39.Rcd1... But Black would have played 38...c4 instead.

To win the game, Black absolutely must find a pawn break. This is the only one available. Instead of 42.Rc1, Kg1 might have been better - really marking time.

43.bxc4 was indicated here. 43...Rxc4.44.Rb1! and whites position holds!!!! 43.Rdd1 was the error.

Does it? I indicated a similar idea earlier, but its success is predicated by White's having an extra tempo. After 44.Rb1 Rxa4 45.Rb7ch it seems to me Black ought to be able to escape the checks. Mind you, this is the first line I would have looked at.