How about this: Duplicates of the old posts are created on a new host using some import/export mechanic, but they also contain a cryptographic signature proving ownership of the old posts (utilizing some secret only the original poster would have access to).

When a reference to a post becomes invalid because it's underlying host is gone, the other host holding those references tries to find new duplicate posts with the correct signature, and if it finds them, updates the broken old references to point to these new duplicates. If it doesn't find any duplicates with a matching signature, it leaves the references broken and has some cooldown before it tries to fix them again.

Considering the system has a way to federate a server self destructing telling other servers to delete its posts could a similar system not just tell other servers to update their URL for that users posts.

Not pretending to know anything about the system feel free to point out why this wouldn't work. I'm just approaching it from a user perspective that having to lose posts to move server is bad for users and having to choose to nuke all posts in the case of shutting down a server is bad for admins.

Like usual, the Mastodon people have the least understanding of how federation actually works.

When a server stops federating, the federated posts still exist in the DB of all the remote instances that recieved it. They're still indexed and searchable from there. You could move those posts into the DB of a new server, but they'll still "originate" from the old one. But you can carry on and interact with old threads.

There's no real foolproof mechanism for updating posts, although there is an AP Update activity, much like deletes there's nothing forcing an instance to respect it.