The process you're referring me to is very much a migration, so you've confirmed the net is that I'd be doing a cross-forest domain migration.
I understand AAD Sync, which we're doing now, gives me a head start. But I'll have to rebuild Group Policy. I'll have to rebuild ACLs (unless I get SID History with this...although relying on SID History indefinitely is problematic, too) and join all the computers to the new domain. And the password change Ken mentions. Fortunately, no Exchange is involved.
So the analysis in my original post was correct: "I can't just add AAD DS DCs to my existing AD DS domain & forest, transfer FSMOs and demote the existing DCs...I have to do a domain/forest migration to a new AAD DS domain/forest."
I've done enough cross-forest domain migrations to know it's expensive and disruptive. The AAD sync headstart is even less of a headstart than I'd get with ADMT!
[Edit] I also note with interest that in the screenshot of GPMC in the AAD DS documentation, the Group Policy Modeling node is conspicuous by its absence. That's an essential tool I'd miss. Though, by itself, it is not a dealbreaker.
But I also realized there's no SID History without ADMT and a Trust, so, yes, I'd have to rebuild ACLs, too.
I will give them a rough estimate of the $1000s it will cost them to "save money" with AAD DS, and the months of character-building pain and suffering they'll endure to get there. On the plus side, cross-forest domain migrations--WHICH IS EXACTLY WHAT THIS IS--are complex and interesting IT projects I enjoy! But I doubt they'll let me have that much fun when they learn the down side.