There is a Rackspace update; from reading it this hit one of the fluke single points of failure in one of their power system.

Considering how data centres are set up there isn't much they could have done to stop it :)

Looks like human error (during the servicing)