You are right. The variation is no good for black. I tried, but still have no idea what to play after 9.f3.

Maybe 9.f3 Nf6 10.dxe5 Nxe5 11.cxd5 Qxd5 12.e4 Qc6 13.Qxc6 Nxc6 with good piece play for black.

Yeah, this might be a pain for Black (I mentioned the possibility in the notes to Atalik-Sax and Kasparov-Anand).
Looking at it again very briefly:
1 d4 Nf6 2 c4 e6 3 Nc3 Bb4 4 Qc2 d5 5 a3 Bxc3+ 6 Qxc3 Ne4 7 Qc2 Nc6 8 e3 e5 9 f3 Qh4+!? 10 g3 Nxg3 11 Qf2 Nf5 12 cxd5 Na5
Maybe this isn't as bad for Black as I first thought, for example, 13 Qxh4 Nxh4 14 dxe5 Nb3 15 Rb1 Bf5 16 e4 Nxf3+ 17 Nxf3 Bxe4.
13 Rb1 Nb3 14 dxe5 Qe7 15 e4 Qxe5 16 Ne2 Nxc1 17 Rxc1 Nd6 18 Qd4
but this still looks better for White.

The recent Gazza-Mickey clash in the 4 Qc2 Nimzo set me thinking - that went 4 Qc2 d5 5 a3 Bxc3+ 6 Qxc3 Ne4 7 Qc2 Nc6 8 e3 e5. Now there have been goodness knows how many grandmaster games in the last 60 years or so with 9 cxd5 Qxd5 10 Bc4 Qa5+ 11 b4 and so on, and as Mickey demonstrated Black seems to be OK (although what nastiness did Gazza have in mind in the 'official' Sapundhiev-Gysi line, I wonder?).

But why do grandmasters never play 9 f3? This is always dismissed on the basis of an old classic Vidmar-Alekhine, which went 9 ..Nf6 10 cxd5 Qxd5 11 Bc4 Qd6 12 dxe5?! Nxe5 and so on. But after 10 dxe5 Qxe5 11 cxd5 Qxd5 12 e4 Qe6 13 Nh3! (Ivan Sokolov) White is just better, isn't he - two bishops, centre, etc? 9...Qh4+ 10 g3 Nxg3 11 Qf2 Nf5 12 cxd5 is also just good for White so far as I can see.

This idea of Solokov's been out there for nearly ten years now, though, and when good players are steering clear in droves, there's normally a good reason. But I'm blowed if I can see what it is. Any ideas, anyone? (preferably John Emms, but I understand he may have his secrets since he's been known to go 4...d5 himself)