Phony Benoni: It's refreshing to see otherwise reasonable people get so worked up about an historical question that The Rogoff Effect sets in. Usually such queries generate total indifferenceThis is one reason I rarely post about discrepancies between ACB and <CG>. That, and the fact that doing so would prevent me from getting anything else done!
I'm also very hesitant to submit changes to opening moving order. That would require more resources than I have available -- and you can rarely be sure.
But I do keep notes in my files, so here goes.
ACB (July 1914, p. 157) gives the "Tarrasch" move order(<1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.Bb5 Bb4 5.0-0 0-0 6.d3 d6>). The source of the score is not mentioned. ACB has a lot of notes from "The Field" (London) for its St. Petersburg games, and in other reports frequently quoted Dutch newspapers. There may also have been reports or bulletins from the tournament itself.
However, with my limited resources, I can't find a contemporary newspaper account. (It's hard to find Alehine games at this time because you never know how his name is going to be spelled.)
Now, here's where Skinner & Verhoeven can help. What is important is not so much what they say, but whom they cite. If they just quote Tarrasch, we're back where we started. But if they happen to have found a newspaper account, that's gold.
I have only one other resource for the game: a collection of Alehine's game in the "Chess Stars" series, edited by Khalifman. (v. 1, p. 244, game #278). It has the "Petroff" move order given on this page. No sources are cited, but I've often noticed that these volumes are based on the score appearing in ChessBase.
This might also explain the origin of the score in <CG>. The Game ID# indicates a very early addition, in the early days when most of the games came from database dumps. I suspect this is the version appearing in ChessBase, and I generally preder a contemporary source like ACB to a database.
Finally, as for the question of Marshall suddenly switching from 2...Nf6 to 2...Nc6. From the Capablanca match match in 1909 to New Yor 1918 he did generally play the Petroff, though he would occasionally play 2...Nc6 against less-than-top competition. Thing is, going into this game, his seven Petroffs at St. Petersburg had resulted in four draws and three losses.
On the other hand Alehine had played 1.e4 e5 2.Nf3 seven times. Six opponents had played 2...Nc6, and Alekhine had no wins in those games, with four ddraws and two losses. In the seventh game, Marshall had played the Petroff and Alehine had won.
With those stats, Marshall might well have thought it time for a quick switch!
But that's speculation. What would be useful now, as <keypusher> suggests, would be to find a contemporary newspaper account,, and there might be a good lead in Skinner / Verhoecen.
And now you now why I don't spend a lot of time posting about these discrepancies in ACB. This took a full afternoon, and there are many of these in every volume.