And I'd like to point out that this is in regards to a rather standard and simplified system. If the server provides a generalized service, like central servers for smaller operations, then firing the admin means getting rid of the one person with the experience to trouble shoot the system in place. So, in the example above you can quickly remove the admin who provides a basic specialized service with higher tolerances than that in a generalized service with lower tolerances.
I think the original argument is a little goofy, I reboot when I don't need to at times, but only because I can - the system is not providing a live service. In fact failover provides support for this sort of thing and is hardly a step away from proper administration.
My openbsd laptop, basically works under the assumption that the system will be reinstalled, from disk, on release. This is a good thing as it insures proper installation of the new system and makes a proper backup strategy a necessity.