Thanks for the ideas, Dom.

A) I also found the game Pap-Gleizerov,Predeal 2006, which went well for Black. But Pap continued to play this line as White and deviated with 11.Nxd4 Bxd4 12.Rb1 when one opponent blundered a piece with 12...Qa4? but it looks like White gets quite a bit compensation after the better 12...Qc3 too:

[Event "Belgrade Trophy 19th"]
[Site "Belgrade"]
[Date "2006.11.26"]
[Round "5"]
[White "Pap, Misa"]
[Black "Pushkov, Nikolai"]
[Result "1-0"]
[ECO "C02"]
[WhiteElo "2430"]
[BlackElo "2482"]
[EventDate "2006.11.22"]

1. e4 e6 2. d4 d5 3. e5 c5 4. c3 Nc6 5. Nf3 Qb6 6. Bd3 Bd7 7. O-O cxd4 8. cxd4 Nxd4 9. Nbd2 Bc5 10. b4 Qxb4 11. Nxd4 Bxd4 12. Rb1 Qc3 13. Nf3 Bb6 14. Qe2 h6 15. Bd2 Qc7 16. Rb4 Ne7 17. Rg4 g6 18. Rc1 Nc6 19. Rh4 O-O-O 20. Bb4 Be8 21. Qb2 Kb8 22. Bd6 1-0

I really like the moves of the white rook. Your suggestion 11...Qxd4 is not as good as it looks at first sight in view of 12.Nb3 when 12...Qxe5  is impossible because of 13.Nxc5 Qxa1 14.Qb3 with a terrible attack and 12...Qc3 13.Qe2 is good for White (+=).

B) I could not find a single game with 9...Nxf3+ 10.Nxf3 a5 and have little trust in that. We boost White's development and make ourselves a big hole on b5. 

C) 9...Nc6 scored badly, so I didn't look at it in earnest. I think the white knight is well placed after 10.Nb3 because there is no pawn on d4 and the other knight can go to g5 sometimes. But I'll have a look at 10...f6!? (Watson) and 10...Nge7. 

Right now I like 9...Ne7 10.Nxd4 Qxd4 11.Nf3 Qa4 12.b3 Qa5 best. White probably gets sufficent compensation but not more than that. The risk to get outprepared is also low.

If you're that scared maybe better to go 5.. Bd7 when (as I remember at least) 6.. Nh6 works a fraction better. Watson gives some rather amusing (if known) piece sac ideas in the play the French. 

I suppose that white can always 'force' relatively similar play to this via 5.. Qb6 6 a3 Nh6 7 Bd3 Bd7 8 Bc2. Seems fractionally neglected for some reason that.
(white gets the bishops tempi back via the inevitable Bd7 -> e6.).

As for 9 Nbd2, I don't know. Would have to look up what my books have to say. I don't remember Watson being so scared but maybe a few new ideas since then.

I'm not an expert in this line, but I play frequently the proper Milner-Barry (9. NxNd4) and have also briefly considered this line as a viable alternative. The inclusion of an extra pair of Knights or in some cases the inclusion of the moves ...a5 & a2-a4 does not necessarily change the spirit of the gambit.

And talking about spirit, I think that the first lines you give look a bit dubious and computeresque to me and I see no real point for White to voluntarily give away 2 of his/her attacking pieces and also open the h-file for Black to use..

The last game you provide looks to have more purpose as White chases the Queen while developing and also manages to penetrate via b7 into the 7th (so Black has managed to keep the centre closed but White invaded anyway!). It also follows logically from b4 which tries to open lines.

But while I was writing that, I remembered we have had a brief discussion of this system some years ago and our friend TopNotch was not overly enthusiastic:

http://www.chesspub.com/cgi-bin/yabb2/YaBB.pl?num=1171043123/2#2

And I don't see why declining the pawn by 6. ... Nh6 is better, e.g. 7. dxc5 and play in Nimzo's style (strengthen e5 and use d4 as a pivot), or play another developing move like O-O..

So not a great help re your question, just my opinion that you should stir your analysis to 6. ... cxd4 and then how to deal with 12. Rb1.