The problem with this sort of choice is that White (in this instance) is really foregoing any real winning chances. Notoriously drawish as the Exchange Variation is means that the stronger opponent need fear no prospect of losing, and thus can seek out ways of extracting the win.

This is an easily drawish position.

I like the choice, given the circumstances. If White makes no mistakes it guarantees a draw.

I also play the 4.c4 exchange French.

The engine seems to think the position is fairly even. But one already has the uneasy feeling (which often happens when I'm playing stronger opposition) that there is more pressure upon White's game than there is on Black's. Perhaps that's not to be wondered at. White has occupied most of the centre; Black is already well placed for undermining operations.

Qb3!? could have been an interesting way to break the pin.

Qb3?! looks interesting but not promising at all. 13. Qb3 Bxf3 14. Bxf7+ Kh8 15.gxf3(15.Bxe8 doesn't work cause black can continue with 15...Ng4! 16.g3 Rxe8 and black does great.) 15...Re7 and white has created weaknesses in own position for no reason. But maybe going for 13.Re1 would be good choice and I am not sure how black would respond but lets take couple examples: [13...Nd5 14.Qc2 Bxf3 15.gxf3 a6(15...Nxe3 fxe3 is good for white) 16.Bd6 h6 17.Kh1 b5(17...Qf6 IM wouldn't go for this draw line against lower rated opponent 18.Be4 Rxe4(must hit otherwise white gets brilliant position due g-file.) 19.fxe4 Qf3+ 20.Kg1 Qg4+ 21.Kh1 leading to draw as white would do better if black plays something else.) 18.c4 Nf6 19.cxb5 axb5 20.Qc5 Red8 21.Qxd6 Rxd6 22.Bxb5 Nxd4 23.Bxd4 Rxd4 24.a4 with most likely draw to follow.] and [13...Re7 14.Qd2 Bxf3 15.Bf4 Qd2(15...Rxe1+ 16.Rxe1 Qxa3 17.gxf3 a6 18.Bxc7 b5 19.Ba2 b4 20.cxb4 Nxb4,now white has nice tactical move, 21.Bxf7+! Kxf7 22.Bd6 Qxf3 23.Qxb4 with draw to follow.) 16.gxf3 Rae8 17.Rxe7 Nxe7 18.Re1 c6 and white is doing just fine.] so maybe IM had some other plans against 13.Re1.

But observe that the pin has renewed itself (in slightly different form) at f3.

Nd2 with the idea of taking up on c4 and later on e5 may have been more beneficial. As there seems to be a big issue with the knight on f3. Feels like white is down three pieces after Qb3 since both the f3 knight and the e3 bishop is pinned to the e2 bishop.

First move I consider dubious. White underestimates the pinning problem and the potential problem in the e column and tries to make stronger the queen side, which doesn't need reinforcements just yet. I think 14. h3 Bh5 15. Re1 would keep the position even. Once you've decided what you're looking for (in this case, a draw) it is better not to get fancy! Plus, concretely Qb3? allows for Black's building of a save path for the pawn advance by playing Na5 later on.

with regards to 15.c4? As stated before, white overlooks the problem in e column. 15. Rae1 was better.

Normally, I like central majorities such as White has here. But such is the pressure, not only against e3 but also c4, that White's King has to come to the rescue...

The King's intervention was just not enough. This is a very instructive example of how to undermine and destroy a pawn centre. Most impressive.

Not to be annoying but, again, 19. Rae1. I think this mistake is the key one.

Interestingly enough, the GK engine much prefers 22.Qa2. I infer that the counteraction against f7 justifies the engine's choice. Observe that a piece behind for two pawns, Black's pressure upon White's centre is stronger than ever.

I agree. Qc2 is not the problem, you sometimes get this in gameknot's analysis, it can look like it was the move that lost the game. But Qa2 made no real difference. Same happens with the rest of 'blunders' gameknot spots from here on. The position is already too complicated for white.

Black's 24th move might also count in favour of 22.Qa2. That array of passed pawns on the Q-side look formidable.

Now 30.Qxa7 c2 looks unpleasant: 31.Qa1 Qd2!? 32.Qxe5 c1=Qch gives Black a big lead (33.Kg2 Qg5ch).

No perpetual can be had here it would be better for white to sac the bishop for the b5 pawn. This way they cannot be connected anymore though things are pretty bad at this point. Qa2 looks to be fairly promising in this set up. Only problem is that it allows Qe3 which could turn out deadly but after the Bxg6 the pressure may be enough for white to form a draw. I think white would be best if she could trade the bishop for the three queenside pawns but it would be very hard to do so.

In my view, worth a try, even if it is not objectively the best. It seems to offer a practical chance at least. unfortunately the Queen alone is not agile enough to stop the King escaping into shelter on the Q-side.++++++++++ yoshark's comments are worth thinking about. But 35.35.Bxb5 isn't really viable, as after 35...Qxb5 36.Qxc3 Qb3 is decisive: the a-pawn is just about home free. For instance: 37.Qf6 a3 38.Qd8ch Kh7 39.Qf6 Qb1ch 40.Kf2 Qb2ch wins easily.

Desperate times call for desperate measures. Nothing to oppose to the move.

Actually, after 35. Bxb5, White doesn't even get 36. Qxc3, as 35...Qxb5+, White must move her king. However, after that, 36...Qb3 is probably still just as effective. An imprecise move on black's part could have left a good chance for perpetual check.