The traditional need for a distributed database is what drove the design that allows a single email domain to contain addressees spread across any number of Citadel server nodes. But as you correctly point out, that kind of thinking is rapidly becoming obsolete (except in the Microsoft world where they maintain that illusion to sell expensive server licenses and even more expensive industry certifications). And no one is going to use Citadel (or Exchange) as the basis of a hyperscale email system, so it's really pointless at this stage.
That's why in the next major version I want to abandon the concept of network-wide email domains. To the best of my knowledge, nobody is using Citadel in that way, and it really just creates a support headache.