No matter how many redundancies, clustered servers or nodes you have, they will break eventually. No system, server, operating system or program is 100% error free all the time.
Stuff is going to happen, so deal with it. Anyone who places all their eggs in one basket will eventually be disappointed. If you do not have systems and processes in place to deal with the loss when they are not available you are simply dim and deserve what you get.
From a personal perspective; Anyone who cannot live without BBM texting for three days needs to get a life.