Messiah: Why 11...♕h4? What is this ♕ doing there? I really don't know what Euwe was thinking during the game, the more natural 11...g6, or the tiny bit more strange 11...♗e7?! are certainly better, at least they don't violate too much opening best practices.
The immediate 13.c5 is wild, but white is not developed enough to acknowledge this move as timely. It is interesting that ♙c5 options will be considered pretty rarely later in the game.
Maybe 13...♗g7, I really don't know. I am sure Euwe already understood here that black is worse, and this is why 13...♖g8 was chosen: he tried to establish complications, in hopes that he could outcalculate Kashdan. 13...♘bd7 is also possible. After 13...0-0-0? 14.♕f5+ leads to an almost winning position for white.
14.♘c3 is also not bad.
The digital beast thinks 14...♘bd7 worth a try: 15.♕b3 ideas lead nowhere, as black is quicker on the kingside, so maybe 15.♕f5 is the best attempt to maintain white's edge, with an incoming ♘e4. I am very curious what an IM or a GM thinks about this position, for me it is pretty unclear how to play properly with black in the long term. The machine also proposes 14...O-O-O 15.♘f3 ♘xf3+ 16.♗xf3 ♔b8 17.♗g4-style of lines. I am not leaving it here, crunching the position for hours or days as I would not understand it's decisions, moreover, I am also unable to calculate it on my own. The eventually played 14...g4 is not a blunder, either.
15.♕e4 leads to chaos. It is possible that objectively it is somewhat better, I am far from being certain.
15...0-0-0, 15...♘bd7 or 15...♗e7 are also considerable, but it requires a very deep analysis which one is black's best attempt to equalize - or better, to go for the ♔!
16.♘f3! is the best. For 16...♕h5 17.g3! will come, with an eventual ♖h1. 16...♕f6 17.♕xh7 ♕g7 18.♕e4+! is also better for white.
18.♗d3 is very interesting. After 18...♘f6 19.♕e2 ♖xg2+?? does not work: 20.♔xg2 ♕g4+ 21.♔h1 ♕h5+ 22.♘h2 +-. 18...♘d7 19.♕xh7 ♖h8 20.♕xh5 ♖xh5 offers active play for white.
19...♗f6 is considerable, black must seek for increased activity.
The silicon monster says 20.g3 ♗g5 21.♕xf5+ ♔b8 22.♘h4 is the best option, with advantage for white. It is not that easy to see why suddenly black's ♗g5 is playable, probably the f5-c8 diagonal is much more important than I think.
21.♘d4 was the best, ♗h4 22.g3 ♗f6 23.♗xg4 ♖xg4 24.♘xf5 defends everything. Soon ♙c5 comes.
21...♗h4 is visually so good that it must be the best move in this position. Black is better.
22.♘f3 was either far or just minimally better. It moves the ♘ towards the centre (and, possibly, towards the queenside to support a pawn steamroller), from d4 it will be able to pressurize the ♙f4, and also discourages 22...♗h4.
Without too much explanation, the engine - after more than 16 hours of analysis - goes for 22...♗h4 23.♘xg4 ♖xg4 24.g3 ♖xf4 25.gxf4 ♖g8+ 26.♔f1 ♕g6 27.♔e2 ♕g4+ 28.♔e3. Black's future is brighter, although the position is far from being won.
23...♗f6 24.♗xf5 ♕g7 is a really cool way to continue the game. It requires some coffeehouse mood, as white's ♙c5 comes, with a possible exchange sacrifice on a1. 23...f4, Euwe's choice, wasn't bad either.
An idea is 24...♕g7. For this, 25.♕e4 is not proper, as 25...♗g5 might hold, certainly 26.b4 leads to an interesting race. 25.♗h3 is the best instead, and white is still better, but here black can escape the ♔ to b8. For this 25...♔b8, the natural 26.♗xf4 ♗f6 27.♖ab1 ♖df8 is a possible continuation. I think 24...♖df8 25.♗e6 ♕g6 26.♕e4 ♖g7 27.♗h3 ♗g5 and 24...f3 25.♗g3 (25.♗h3? ♕h5! ∓) are playable.
25.♗e6 was more precise, as 25...♕g7 runs into 26.♕e4, and black's attack goes nowhere.
25...♖h5 was a possible way to continue the attack, although 26.♕e4 ♗g5 27.♗e6 ♕f8 28.♕d4! (and ♙f4 is undefendable) favours white, but black is at least not dead lost. 25...♖g7 is sad, but probably the most precise, heading for a long and difficult defensive work. The point is that g5 should be freed up for the ♗, as we have seen it in previous lines.
26...♖g7 was necessary, however, after 27.♖e3 ♖f8 28.♖ae1 white is obviously better. I've got zero idea what Euwe thought here, possibly he was in a serious time trouble. The rest is a total collapse.