It depends a lot on individual departments. Most astro groups I've been in have had very competent sysadmins who aggressively lobby for fault tolerance and assu

It depends a lot on individual departments. Most astro groups I've been in have had very competent sysadmins who aggressively lobby for fault tolerance and assume that the university proper is incompetent (the very best kind of angry greybeards). So all the server rooms have enough UPS capacity to gracefully kill servers, data is properly backed up and so on. At some places like MSSL we even had an onsite diesel generator becuase there were flight critical experiments going on all the time. Many other faculties don't need that level of resilience so they don't bother. I'm not sure what the situation is for wider engineering or bio/chem. I imagine there are serious health and safety concerns for things like fume hoods that must be able to evacuate gases in the event of a power failure.