It looks as though Black might have 'won' the battle of the databases. From here on, the database indicates good results for Black. But the numbers are very small, which suggests plenty of room to test unexplored options.

The alternate option was 15...Bd7, which was chosen by the higher rated masters. The results are otherwise statistically insignificant.

Instead of cxd5, Bf8 is a move in the Gameknot opening databases, as in Sobek, Jaroslav (2395) vs. Jaworski, Miroslaw (2245) which ended in a draw, or h5 as in brettbevans (2453) vs. digitron64 (1959) which black evntually won. We seem to be out of the Gameknot opening databases now.

This seems to me a reasonable move to consider, as although White gets a passed pawn it is isolated. I'd be very concerned, though, that it would be very difficult to attack on d5, as Black's bishop can not reach it, but White's can defend it. That would tend to leave the White knight with a free hand (once a rook gets on d1) to play to e4, reaching into the rear areas at f6 and d6. The 'bishops of opposite colour' can sometimes be an aid to a defender, but in different circumstances benefit the attacker. It has the effect of putting Black a piece ahead on the dark squares, and a piece behind on the light.

Agreed. The following exchange of knights enabkles White to establish his bishop on a fine central square, targeting the b7 and f7 pawns.

There is not much that can be done about White's plan to double rooks on the c-file. I'd probably play 21...Bf8, here, with the idea ...Bc5; ...Be4 - also a fine outpost square for the bishop. By doubling rooks on the c-file, White will more than likely win the b7-pawn, but after that his task would become, I think, difficult and highly technical.

It might be something to bear in mind. I still think ...Bf8 is playable even after White's next. Another plan here might have been 21...Re8, with the idea of doubling rooks on the second rank. This is admittedly a passive approach, after which Black then will have to think how to bring the bishop into effective action, but such a fortress-like defence might not be so easy for white to break down.

Maybe to move it to c5 and then maybe d4

I think that is a very reasonable plan, as I said earlier! It seems to me playable yet.

The pin on f7 is mildly annoying, but not yet dangerous. Preferable seems 23...Bc5 24.b4 Bd4 and if 25.Rc7, Kf8 26.Bxb7 Ke7.

Yes, but it still takes some winning for white (or maybe that should be some losing for black?)

True, that White bishop is a bone in the throat, but Black should be thinking in terms of defence, by now, as White has the edge, though material is even. No harm in keeping an eye out for your opponent's mistakes, of course. I agree with charleshiggie: Black isn't losing yet!