If you have to ask that question, you clearly haven't been tracking their up-time statistics closely enough. *grin*

If you have to ask that question, you clearly haven't read any of the architectural discussions or seen any of what I've said on the subject in public forums. Or what Twitter's said as far as thinking about the company as a "communications utility". ;-)

The internal infrastructure is all PubSub, more or less, and Twitter exposes a PubSub interface, but only for public updates, and only by request. The infrastructure to do more is there. The hard part is that it *is* a website, and displaying low-latency, high-volume data is hard. Not hard enough to justify the current downtime, but it's still non-trivial.

FWIW, I got someone at Google to admit that AppEngine wouldn't work for Twitter, at least in AppEngine's current incarnation. There are ways to do it, but you'd have to have a system external to AppEngine.