Need analyis, Poisoned Pawn Accepted gone wild

I just finished a ridiculous game.  My personal greatest attack, I think.  It was in a line of the Najdorf: Poisoned Pawn Accepted that has been thus twice refuted OTB and I managed a remarkable win after multiple piece and pawn sacrifices.  The question is.  How unsound was the play from move 18 on?  Can I get some serious thoughts on the issue, I've been a fan of this triple pawn sacrifice line but I am not sure its viable after this game, I suspect I got a bit lucky.

Sleight you would have completed in 31st step itself. If you move the Queen to a3, the game would be over.

That is not the case because of the defensive resource 31 ... Qa4 that is used here and then mate actually fails, because nothing prevents the King from escaping back to b6.

But 32.Qa3+ Qa4 33.Qc3+ b4 (33...Qb4 34.Ra1#) 34.Qc5+ Qb5 35.Ra1# is a little faster.

Then to the position after move 18 -- is this theory?! I knew this line was insane, but does white really have enough there for two knights?

What happens after 28...Qe4 instead of Be4?

This is theory, or so I have been told.  Whether White has full compensation is what needs to be proved.  That's why I need suggestions for Black and White's play.