After 1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. Bg5 e6 7. Qd2 a6 8. OOO Bd7, both GM Kozul and IM Kanmazalp devote many more chapters to 9. f4 than to 9. f3. I think the reason is that 9. f4 has been around longer and also lends itself to more concrete analysis.
After 9 f3, both books discuss the relatively recent 9. ... Nxd4!? 10. Qxd4 Be7. This scores very well for black  46%! Playing this may not be so easy as the stats indicate. White's plan is simple, a pawn storm on the king side. Black may have to choose judiciously between playing on the queen side, king side, or center.
I have Kanmazalp's book and he now covers 11.h4, 11g4, and 11. Kb1. White can easily combine these moves, so there are many possible transpositions.
I played through his analysis of 11h4 with a computer and found the engine to be almost worthless. The line in question is a mutual pawn storm and the engine simply can not make out the strength of attack until it's almost over. I'll just say you get everything you hoped for in a Sicilian.
I got bogged down looking at 11g4 because black must consider Be3 g5 by white. So where to put the knight? Kanmazalp opts for h5. If black plays ...Bc6, then ...Nd7 becomes possible. ...00 makes ...Ne8 possible.
I found only 17 games after 1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. Bg5 e6 7. Qd2 a6 8. OOO Bd7 9. f3 Nxd4!? 10. Qxd4 Be7 11g4. Black scores 50%. Tal played white in the first two games (a win and a draw).
After 11g4, several moves have been tried: ...b5, ...Bc6, ...h6, ...h5, ...Qa5, and ...Qc7. At this point, I don't think I can trust game results or computer analysis, without digging deeper.
After 1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. Bg5! e6 7. Qd2 a6 8.OOO Bd7 9. f3 Nxd4!? 10. Qxd4 Be7 11. g4, Kanmazalp continues ...b5 12. Be3 OO 13. g5 Nh5 with further analysis of 14Qd2 and 14h4. 14f4 is also possible.
