Rick Kennedy's well-researched Alekhine vs. Marshall's 1.d4 d5 2.c4 Nf6!? at ChessCafe (originally in Kaissiber #27) would almost lead you to believe that Frank James Marshall's center-surrendering experiment against the Queen's Gambit might be fully playable. Alekhine's notes on Alekhine - Marshall, Baden Baden 1925 suggest as much, and 12…Nxe5 13.0-0 0-0 14.Be2 Be6 would clearly have improved on Marshall's play. However, there are two lines that Kennedy does not consider which seem to keep the line in doubt, and both are examined in Matt Pullin's excellent two-part video series from 2008 (see above).
I always admire Pullin's objectivity, and he does his best to demonstrate Black's chances as well as White's most powerful challenge with 1.d4 d5 2.c4 Nf6 3.cxd5 Nxd5 4.Nf3! (objectively better than the natural 4.e4 Nf6! 5.Nc3 e5!) 4...Bf5 5.Qb3! and if 5...Nc6 6.Nbd2! gives White a strong variation of the Baltic (1.d4 d5 2.c4 Bf5), as demonstrated in the game Takacs-Havasi, Budapest 1926. Kennedy also does not mention the game Alekhine - Mooyman/Citroen, Surabaya 1933 (surprising, given his focus on Alekhine) where White was definitely better following 1.d4 Nf6 2.c4 d5 3.cxd5 Nxd5 4.e4 Nf6 5.Nc3 e5 6.Nf3 exd4 7.Nxd4 (7.Qxd4 is also slightly in White's favor). Pullin suggests Black may have a playable game here after 7....Bc5 8.Be3 O-O (better than the tempting but tempo-wasting 8...Ng4?! as Alekhine's opponents tried) and if 9.Ne6 Bxe6 10.Bxc5 (gaining the two Bishops in an open position) 10...Re8 11.Qxd8 Rxd8 he thinks Black has reasonable chances, which may be true, though Black's position is hardly inspiring. Conclusion: Marshall's variation of the Queen's Gambit Declined may have more to it than commonly thought, but it does not inspire confidence against White's best counters.