I'll chime in...
100 copies the first year, 20-30 copies the following years is what you can expect from a really good game. With the overall number of available homebrews constantly increasing, those numbers are actually more likely to shrink than to rise.
That sounds about right. IIRC, Fall Down sold 60-something copies in its first few months.
As others have said, it's on the order of hundreds. I could see a really big project, say an RPG of some sort, going over 1000 if done properly. I would also think David Crane's estimate of 1000 hours is accurate, given that working in a corporate environment tends to slow a person down (e.g. from working when you don't feel like it).
To be honest, I'm an attention whore, so yeah, when one of my games does well I tend to write another. Royalties are a plus, but they're not significant enough to worry over. That said, I've found developing for the Atari to be more profitable than for say, the Palm.
Personally I'm in favor of freely available binaries... from a small-time developer's perspective, they increase word of mouth, if nothing else.
Generally I think that would be irritating, though I don't actually own any of the first-edition games in question.