Or practically a million other lines...

Actually, If Black wanted a Dragon. 2...d6 is a more reliable way of getting there. Meanwhile, this move leaves open for Black a wide range of possibilities, an 'Accelerated Dragon' being one...

By playing 3.Nc3 first, White was being cagey: would he play the Closed Sicilian? With this pawn advance, White signals his willingness to enter the Open Sicilian after all.

Yes. 8.Bxc6 now comes into serios consideration, breaking up Black's Q-side. Black really ought to have taken on d4 (4...cxd4) transposing into normal Open lines.

10.Ne5?? losing a whole piece.

There was a way of playing this that led to an edge for White: 10.Bxc6 bxc6 11.Ne5 threatening 12.Bxf6 and winning a piece. What a difference move order makes.

A reasonable approach. Going over to the defensive is a sure fire recipe for a loss. Aggression might yet turn the game around.

Better is 12...Kc7.

So 12...a6 was far-sighted enough to save the piece. But paulmorphy91 is correct: 12...Kc7 would have been an improvement.

if preparing a pawn storm why not 18.g4?

It does tend to write down the doubled rooks. I would have taken with a rook, and then brought the king up to c4.

This isn't a bad bishop, black is still winning by advancing the kingside pawns.

Yes, technically, Black's is a 'bad' bishop, by which is meant that most of Black's pawns are interfering with its mobility. However, it has to be said that 'bad' bishops are very occasionally more useful than 'good' bishops. For the time being, Black's prelate is fulfilling a useful defensive function, and his pawns are mobile enough that the 'bad' bishop will eventually be on its very best behaviour.

now, white is winning.

Black completely lost the thread - easy to do when events take an unexpected turn. Possibly a better chance lay in 32...Rd1 (which suggests White would have been better off playing 32.Kd2).