sagahelten: Here's the analysis by Byrne from New York Times:
Bronsteinís 4 Nc3 introduced a sharp gambit, in which Black has trouble even if he refuses it. If 4 ... e6 5 e4 Be7 6 Bc4, White has the superior pawn center without being obliged to pay anything for it.
After 9 ... f6, the white knight was denied the invasion squares at e5 and g5. [...]
This move 10 g3 was tried out repeatedly after this battle and is still the subject of debate. At first I thought Bronsteinís fianchetto was the right way to go. But I later learned that my old friend Al Horowitz, a star on Olympiad teams in the 1930s and later the chess columnist for The New York Times, had discovered an even more dangerous weapon in the weak-looking yet powerful 10 Be2!
After 12 Be3, it was necessary to get my queen off the semi-open e file as soon as possible, but 12 ... Qc8 13 d5, threatening to open the game before I had finished my development, was dangerous.
Whiteís 15 ab was wrong because it opened the a file before he was ready to exploit it. He should have considered 15 h4 with the idea of 16 Kh2 and 17 Bh3. I am not sure I could have defended against an incursion on e6.
If 35 ... Rc5, then 36 Ra8 would have won for my opponent.
But 35 ... Re5 virtually finished the struggle. If 36 Qc6, then 36 ... Qc6 37 Rc6 Kf7 38 Rc8 Ke6 39 c6 Kd6 40 Kg2 Rc5 wins.
So, hoping against hope, he played 36 Qh3. But after 42 ... Rd5, he resigned.