"Master" generally refers to someone who has received a FIDE title (FM, IM, or GM) or met the criteria for becoming a master in one's home country (usually a national rating of 2200 or higher). So these players are not the very best in the world and usually not GMs, but they are still well within the top 1% of serious tournament players. Welling, Grund, and Schlindwein are 3 players rated above 2400 that have played Na3 in the last 10 years. Grund played it against another near-2400 opponent in 1996, and the game was a draw. By the way, those 26 games are in a database of over 2 million total games, so yes, Na3 is still very rare. This database is Chessbase's Megabase 2003, in case you're interested in finding out more about it.
However, often even strong grandmasters don't have very rigid ideas about opening moves. Pavel Blatny is an American GM who regularly plays openings like 1. b3 2. Bb2 3. g3 4. Bg2. You can see him in action if you ever visit a major tournament in the US!
Also, I don't believe that preventing Nc4 should be a motivation for black's move choice. Instead, I feel black should welcome white's knight to c4, whereupon black will soon chase it away by playing d5 or b5 (with Bb7 to follow), or perhaps even simply leave it there to block white's bishop from accessing c4. A more harmonious plan for white seems to be to play c3, Nc2, and make a strong point at d4 by advancing the pawn there. I do agree, though, that 1... e5 does not threaten to play Bxa3. Indeed it's rather difficult for black to threaten much of anything in the first few moves if white doesn't place his pieces aggressively. But my motivation for choosing my opening moves is to place my pieces and pawns in such a way that they work well together, not to make a threat that I expect my opponent will see and adequately respond to.
You're right that my moves in that game are almost all bad until 11... hxg5?? It's intentional. The whole point of the early play is to (unsoundly) give away material (which computers love to take) to remove black's defensive pieces from the kingside (as Qxa3 does). And the idea behind the later play (putting the knight pointlessly on g5, a square on which it can do nothing if black simply ignores it) is to tempt black to take more material to open the h-file. What makes the h-file attack dangerous to a chess engine is that the mating moves (bringing the heavies to the h-file) are not checks or captures. When it evaluates capture or check, a chess engine will extend its search until the captures and checks end, thus seeing many moves farther ahead. Since the threatening moves are not checks or captures, it sees the danger too late. Of course, playing this way against a human is ridiculous, because humans think much differently :)
I don't suppose you play online at FICS (I'm toddmf there) or ICC? Maybe we can meet online and play or talk chess sometime instead of clogging this video game forum with chess stuff :)