Recently had this move played against me online. By keeping the pawn on g5, Black tries to solidify the kingside so he can develop his queenside bits and castle long. Since it was 3 minute blitz, i didn't react properly and after 1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 e6 5. Bg5 h6 6. Bh4 dxc4 7. e4 b5 8. Be2 g5 9. Bg3 Bb7 10. h4 Rg8 11. hxg5 hxg5 12. Qc2 Nbd7 13. a4 a6 14. Ne5 Qb6 15. Nxd7 Nxd7 i think White was already worse!

There's no mention of this in Cox's book (or any book), and found only one game in the database of any significance: 

[Event "Acropolis op"]
[Site "Athens"]
[Date "2006.08.13"]
[Round "4"]
[White "Al Sayed, Mohamad Naser 2492"]
[Black "Karayannis, Athanasios 2380"]
[Result "1-0"]
[PlyCount "55"]
[EventDate "2006.??.??"]

1. d4 Nf6 2. c4 e6 3. Nc3 d5 4. Nf3 c6 5. Bg5 h6 6. Bh4 dxc4 7. e4 g5 8. Bg3 b5
9. Be2 Bb7 10. h4 Rg8 11. hxg5 hxg5 12. Qc2 Nbd7 13. Ne5 Nxe5  14. Bxe5 Nd7 15. Bg3 Nb6 16. O-O-O Bd6 17. Bxd6 Qxd6 18. e5 Qf8 19. Ne4 O-O-O 20. Rh7 Kb8 21. Qd2 Nd5 22. Nc5 Nf4 23. g3 Ng6
24. Qxg5 Rh8 25. Nxb7 Rxh7 26. Nxd8 c3 27. Nxc6+ Kc7 28. Nxa7 1-0

So maybe 10... Rg8 is a novelty and a start of a new trend?!  Anyway in this game White simply castled long (that's news to me in the anti-moscow, although in some lines of the Botvinnik White castles long too) and then probed Black's kingside. Black blundered a pawn with 23... Ng6?? but White already had some pressure. Is this approach best for White?