Twitter's certainly having a lot of trouble, lately. Today, apparently, their main database had a bit of trouble. Now, first of all, I am a little surprised that a service on Twitter's scale has a main database, as such. What does it do? I'm also very surprised that it was allowed to reach a state where it had too many connections; given that they're using Ruby on Rails, they should, I believe, have control over how many worker processes there are total, and thus be able to set this to something that their database can manage.
Beyond that, though, am I the only one a little surprised that they're not using simple, boring ol' MySQL Master-Slave replication? It's not rocket science, and should probably save them in the case where the master dies; just promote one of the slaves.
You have to wonder just how competent the implementers are, really.