At last, after an age of lurking, a subject where I do feel qualified to comment. In my organisation the operation of data rooms is split between IT and Real Estate - I am an engineer working for the latter and am responsible for all M+E services supporting the equipment. (Anything inside the racks belongs to IT).
For highly critical loads there is always more than a single UPS supplying critical power - at least 2 and more usually 4 or more. Cooling is also on a redundant backup system, details vary depending on the method used, ditto generators etc and we even consider water buffer tanks. Technically these are 2N services with distributed supplies and need to be fault tolerant AS A SYSTEM. There is also a similarly resilient backup data room in a distant location.
There's always scope for idiocy, but any work is controlled by a comprehensive change control system with at least 3 individual approvers before it's authorised. I can't comment with the same authority on the software side although an element of resilience is always included, and failover procedures are regularly practiced by IT. At least once per year I fail various elements of the support systems to test the backups and stress test the whole installation.
There is absolutely no way a well designed and maintained IT infrastructure could fail so catastrophically from a power irregularity - this is bread + butter stuff. An installation that is not well designed or maintained however.....