A couple of days ago I started looking at 1.d4 Nf6 2.Nc3 d5 3.Bg5 Nbd7 4.a3?!. (Dia)

4.a3 appears a silly move but after 4...c5 5.dxc5 it’s not at all clear that Black can win his pawn back. The position arising after 4...e6 5.e4 (5.f4 transposes to Steward's line) 5...dxe4 6.Nxe4 Be7 7.Nxf6+ Bxf6 8.Bxf6 Qxf6 (Dia) must be worse for White than similar French lines (Burn and Rubinstein) but may still be somewhat easier to play for White:

a) 9.Qd2 c5 10.Nf3 0–0 11.0–0–0 cxd4 12.Qxd4 Qxd4 13.Rxd4 += Schinzel-Pinkas, Bydgoszcz 1976.
b) 9.Nf3 0–0 10.Qe2 (10.Be2 e5! is at least equal for Black) 10...c5 11.0–0–0 cxd4 12.Rxd4 e5 13.Rd2 Re8 14.Qe3 a6 15.Be2 Qe7 16.Bc4 h6 17.Re1 += Schweber-Quinteros, Villa Martelli 1996.
Whether you find such positions attractive or not is to some extent a matter of taste. From a practical viewpoint it must be taken into account that some black players may dislike them. Quite possibly Black must look into the untested 4...c6 or 4...h6 if he is looking for a more interesting path to equality.
Rubbish? I honestly don't know, but now I can with a clear conscience exclude these lines from my manuscript!