That way we have resilience against single points of failure for user accounts, we can pool admin resources, while still being able to host our own instance, do our own thing if we choose. Admin teams can be made up of the admins of every (or a subset of) connected node in the foo.xyz clustered instance, and if one or two admins want to bail, the network stays up.
There's still issues over the domain name ownership, load balancing, etc., but still. Just a thought.
This is Kithop's private, invite-only instance - hosted on FreeBSD in Canada.