Also, a site like Reddit wants something like 99.9% availability: roughly 8 hours of downtime per year. Lemmy instances are probably satisfied with 99% availability: roughly 3 days of downtime per year. If one instance is down, but the rest of the fediverse is up, it’s a bit annoying, but not devastating. Users of that instance might have to create alt accounts on another fediverse instance, and certain communities would be offline for days. But, as long as the entire fediverse itself doesn’t go down, it’s not the same as a Reddit outage.
Getting that extra “9” of availability means having engineers on call, it means having a technical staff that creates and maintains monitoring systems, does capacity planning, runs disaster preparedness scenarios, etc. It’s expensive.
Some fediverse admins might run monitoring systems, either because they really care about their instance, or because doing it is interesting and fun. The ones that don’t might just have to do reactive maintenance when something breaks. But, because you’re only aiming for 2 nines, it doesn’t have to be a full time job.