I think Black's opening play was fine.  With colors reversed, Nunn calls 6.Bd3 "a harmless attempt to avoid the main line", but doesn't show it too be bad either.  I think White's extra a3 move is fairly insignificant and it's safe to say the position is equal.

First of all, I mistakenly said that McGrew's articles appeared in his Chesscafe column "Gambit Cartel".  In fact they appeared under the "Skittles" section at that site.  Go to http://www.chesscafe.com/archives/skittarch.htm and look under Giuoco Fortissimo: The Rousseau Gambit by Tim McGrew, parts I and II.  That should give you a very good idea of how to play 3.a3 Bc5 4.f4 as White (Most of the time a3 is of no use whatsoever.  It just prevents the most dangerous variation.)

I can't really remember what is best for Black after 3.a3 d5 4.exd5 Nxd5 (4...c6!?) 5.Qh5.  There have been 2 whole books devoted to this line from the Black side and I'm sure you'll find at least a footnote in NCO.  I'm pretty sure a3 isn't always useful.  I believe it is supposed to eliminate the most dangerous tries (actually I haven't looked at this in some time, if someone were to actually try this then I'd suggest they look into the theory in more detail).

After 3.a3 Nc6, Myers recommended 4.Nf3 (as you suggested) answering 4...d5 with 5.Bb5 I think (I can't find my copy of his book!).  I remember also looking at 5.Nxe5!? (see "101 Chess Opening Surprises" for some analysis with colors reversed).

I only used this line in "3 0" games on ICC.  I had never planned to use it as a serious tournament weapon I don't suggest anyone else does either!  I do remember having some fun trying to find the best ways to use the extra move a3 (it's rarely harmful).  For example my move versus the reversed Ruy Lopez try was very effective but you'll have to find that for yourself.

LOL!  You really need to learn how to erase better. 

A reversed 2 knights defence is also what Myers recommended.  I'd agree that after 1.e4 e5 2.Nc3 Nf6 3.a3 Bc5 4.Nf3, that 4...d6 is the best move (4...Nc6 5.Nxe5 is usually rated equal with colors reversed but I always felt more comfortable being the one to take the pawn and play d4, or ...d5 rather).

The line that bothered me was 4.Nf3 d6 5.Bc4 (I think White has a hard time proving an advantage after 1.e4 e5 2.Nc3 Bc5 3.Nf3 d6 4.d4, so I really doubt if 5.d4 is anything special.) 5...Nc6 6.d3 Be6 when it seems that having a pawn on a3 is harmful.  Well that's the way I felt all those years ago.  Let me know if you see something decent for White here.

Of course I keep pretending that your 4.Nf3 doesn't fight for the advantage and 4.f4 does.  That's certainly not the case.  I do think it has or more dynamic quality to it though (some detractors might say crazy instead of dynamic  ).   

Oh, and McGrews analysis indicated that White's strongest line (after 1.e4 e5 2.Nf3 Nc6 3.Bc4 f5? 4.d4! fxe4 5.Nxe5! d5!?) was 6.Bb5 which is obviously bad here (6...Bb4 of course).

I think Black's best might be the simple 1.e4 e5 2.Nc3 Nf6 3.a3 Bc5 4.f4 d6! when he may well be better but see McGrews article which shows things aren't easy.

This suggestion is intended more for  blitz or skittles use.  Although I suppose it could be used as an occoasional surprise weapon.  I'm sure most of you follow Tim McGrew's Chesscafe article about Gambits.  If so you will remember a series of articles on 1.e4 e5 2.Nf3 Nc6 3.Bc4 f5.  Eventually he concluded that theoretically it wasn't very good but did have some practical sting.  I'm proposing that White try this same gambit via 1.e4 e5 2.Nc3 Nf6 3.a3 Bc5 4.f4!?.  Many years ago Hugh Myers wrote a book titled, "Reversed King's Pawns: Mengarini's Opening".  The entire book was devoted to 3.a3.  The main idea being 3...d5 4.exd5 Nxd5 5.Qh5! and Black is unable to respond at any time with ...Nb4.  I was never satisfied with his (Myers) solution to 3...Bc5 and put the opening on the shelf for many years.  I followed the McGrew articles with interest but didn't think of it as a serious opening.  After 1.e5 e5 2.Nf3 Nc6 3.Bc4 f5 4.d4! fxe4 5.Nxe5 d5 6.Bb5 Black is in trouble if White knows what he's doing.  Somewhere along the way it occurred to me that I could try some of the Myers' suggestions but try 4.f4 against 3...Bc5 without allowing ...Bb4.  Of course it's nothing your going to see a respectable GM playing but it could throw off the average club player perhaps. Let me know if you try it out.