Git's simplicity and pile-of-tools approach, along with daily use of the tool and prodigious debug output, strike me as exactly the sorts of things that minimize the need for bug tracking. If there's a bug in my own statistical code, for example, I usually know about it from the debug output. If that works for the Git developers, fucking awesome.
However, just because Linus doesn't use a bug tracker, doesn't mean that the big shops (RedHat for example) don't pick up the slack. I think Linus is a flawed example ;-)