Imagine if every tweet included info about where to find that users' status messages if Twitter is down. It could simply be the URL of a feed. Unless Twitter went down it would just be stored by your client. But if Twitter should go offline, at least you know how to find out what's going on with at least some of your associates.
The point is that with annotations we have the ability to create new ad hoc flows of information, and can possibly use it to make Twitter more reliable.
It's like having a place to meet your friends after a ballgame, or part of being prepared for an earthquake (I just landed in SF). You arrange for a meeting place if the normal lines of communication go down. You use the normal lines to transmit the place you will be found in cast the line goes down.
I guess it's something like DNS, where you always specify two servers and two hosts. If everything is working you only need one. But the network is designed to work even if everything isn't working.
Of course it's this line of reasoning that led me to the conclusion that Twitter should just run off DNS, and Twitter should be the Verisign of this network. But that's another story.
Just had an idea that with the new 
