Cokstikyan writes seven rules for succeeding in marketing a massively multiplayer game. The first is Never, ever ship before you're ready.
That's interesting, because the usual wisdom in software today is the opposite: ship early and often. But it's probably the right answer, Iterative design works because it helps you elucidate exactly what the business customer wants, but that may be hopeless in games because the game just isn't worth enough to the customer. Perhaps massively multiplayer games are the last refuge of waterfall design...
On the other hand: Ed Blachman writes to say that my thinking here is probably out of date. In fact, "the software industry" might not exist any more, and it's certainly not one industry; the dynamics of shrink-wrap end-user software are very different from corporate open source, corporate non-open-source, and all of these are different from Web services.