14...e5 is a blunder. Sokolov should've continued with 15. d5 Na5 (15...c4 16. e4) 16. e4 c4 17. a4 Bc8 18. Ba3 19. Rfe8 where White has considerable advantage (passed, supported d pawn, centre and kingside space, pair of Bishops, outpost in f5) He continued with 15. Qc2, and Judit should've played 15...Na5, in order to avoid in part the trouble coming from a White passed d pawn as seen in the previous line. But she played 15...Rd7, where I believe 16. Ne4 would be slightly better than Rad1 (not noticeably though).

Now, 17. h3 was just pointless. The prophylaxis isn't appropriate here, and Sokolov should've played Ne4, with advantage. But 17...c4 seems much worse (takes some of the pressure of the d pawn and helps making it passed) than ...Bb7, backing up the Knight and setting for a discovered double attack after 18. d5?! Up to move 25, it appears to me White has a sizeable advantage, but 25. Qb4 is a blunder, where Qe2 or Qf1 would have been safer. This is really where Judit's attack starts, and on move 30 perhaps Sokolov could've improved by Qf1 (protecting some squares around his King), but White is very probably already lost there.

This attack is actually what leads me to think that something wrong went in the moves 18-25 in Sokolov's play, since I doubt Qb4 was a mistake enough to set off the loss.

Again, this is all handmade analysis, so I'd like someone with Fritz or some other engine to check it for what I missed.

The Nimzo can be treacherous. Spassky had most of those advantages here and it didn't help him much:

Nice attack by Polgar. Wow!! Is she in great form.

In this particular system (Botvinnik), Black induces a dark-square blockade by White, hemming in his own Bishop. Looking at the position, Judit must've concluded that White was going to seek counterplay on the light squares, and therefore decided to blockade these too. One might think this harms her own light-squared Bishop as well, but the fulcral point here is that this Bishop has the open c8-h3 diagonal to control whenever he wants, while Sokolov's has no way at all to reach an open diagonal, after 19...e4 (a logical follow-up which I guessed on my own - weee!).

In the aforementioned Spassky - Fischer game, Fischer played the Hübner variation with ...e5, tempting White into d5 which closes the centre and makes the light-squared White Bishop look very bad! The difference between these two games is essentially the color of the blockade, and the more advanced Black pawns for Judit. The strategic concerns are, however, very similar.