This variation isn't as strong as it used to be but it's still viable. Usually leads into a poison pawn game, at least from what I've seen it's statistically the best for black.

The thought occurs that maybe White should "get in" a2-a3 before Black can play ...b4. What is the current state of theory about such a move?

Doesn't seem to much current theory on this variation, as 11 f3 is almost always played now. John Elms had this to say about White playing a3 in a 6 Bg5 Najdorf ( Neither side had castled yet) "9 a3 is obviously playable, but Black is usually happy when this happens -- White's attack is slowed down by a tempo and Black has more time to construct the perfect defensive set-up." I also would add that I have been studying Nikolay Minev's book "Mastering Tactical Ideas". Minev has a chapter called the "The Notorious Neighboring Square" mostly talking about g6 (g3) after h6 (h3) has been played. His tactical examples have made me very leery of playing a3 after castling queenside. Most of Minev's examples involve sacrificing a piece on e6, forcing Black to play fxe6 and then planting a piece on g6.

Qa4? why not just take on b7? if he hassles you with the rook you've still got Qc6 + or Bc6 + etc to get out of it.

One of my main goals when I started analyzing this game was to look where I might have returned the pawn. After sleeping on this game for a few nights, it is clear how this should have been done. If material had been equal, a wing advance on the kingside would be the correct idea here. With a pawn plus it's even stronger. If I advance my g-pawn (perhaps supported by the h-pawn) I should have little trouble giving back the pawn while breaking up Black's kingside structure and obtaining a open file for my heavy pieces on the kingside.

It's hard to disagree with that comment blake78613. The move ...d6-d5 is not all that dangerous just yet, as 17.g4 (say) d5 18.Qd4 pins Black's d-pawn, threatening to win it, but also attacks g7.

As the game unfolds, this Nc3-move is a waste of time. More logical and efficient seems 18. Rd3 followed by Rhd1. Later on, Rd3 can then be used as a defender along the third row in the same way as in the actual game. But this time there wouldn't be a knight in the way. White would have two more moves to start some kind of attack in comparison to the game.

The move seems unobjectionable, to me. Earlier, the knight was forced away from this square onto an inferior posting at e2, since when, the c3 has become once more available to it. The knight seems better posted on c3 than e2, so the move ought to constitute an improvement in White's position. Possibly the move could have waited until a time more opportune, but I don't see in the move played the seeds of defeat. White is a pawn up here, but Black has the more active game. Probably the game is fairly level at this point.

There is nothing wrong with 18. Nc3. In fact, I judge it best. But that depends upon what plans you have in mind.

Now, i remember. I wanted to leave d1 open for the c3-knight. As mordicus points out above, I wasted several tempi moving the knight around that could have been used to advance my kingside pawns. When I played the French Defense, I won a lot of games running my King to the other wing, but in the French White has a lot more difficulty breaking open the center.

The move played has its point other than the knight retreat, though. White can again think about e4-e5 to close the centre, then launch into his own K-side attack.

Your problem is that you don't have a coherent plan of attack even though you appear to be trying to do so. In fact one is hard to find, so you shouldn't be trying so hard. Look for ways to improve that position without wasting time. I disagree with 19. Rd3. If 19.e5 instead, then White need not be afraid of Bxf3. That would simply improve White's center and give White the lead.

Instead of 21.Nd1, 21.a3 might have been considered to prevent this incursion.

It has to be said, though, that even after 23.g4 Black's attack has a long head-start.

White still has his extra pawn, for which Black has compensation in terms of the initiative and the more active position. But White's defensive resources at this point seem sufficient to put up a strong defence ...

It seems from his note that White overlooked something in playing this move. At the moment, Black has practically no attack, so White has time to organise his pieces. A good centralising move is 27.Qe3, doubling heavy pieces on the e-file (which brings the move f4-f5 into consideration), and also adding protection to the c3-knight. That might allow a sequence like b2-b3, Kc1-b2. At any rate, I'd be inclined to stay in the house until the bailiffs are knocking at the door...

Now Black really does have an edge, I think, but White isn't out of it yet! White might have considered exchanges here, but possibly wanted to keep the central files closed.

Not yet, by any means, though 30.Qe2 was the better choice, here. Black could now have played 30...Qa5, threatening 31...d4. White could not have played 31.Kc1 on account of 31...Be3+, or even 31...Qxc3! 32.bxc3 Be3+ 33.Kd1 Rxb1 34.Ke2 Rxb1 (34...Bb5+) 35.Kxb1 Bxf4 and an easily winning endgame. After 30.Qe2 Qa5?! 31.Kc1 looks fairly safe.

Rather an abrupt finish! White was doing fine - at least an even game - until the 27th move, when White's over-defensive stance - I think based on an over-pessimistic evaluation of his game - really invited Black's attack.