Right but my point is they would just submit the request to the host server. If the original is taken down then all the federated service will lose the comments as well.
Not how federation works. Let's take a lemmy post as an example. If a server is federated with another and a new post is made, all subscribed servers are notified and a copy of the item is sent in that notification. If the original is "taken down" the copies still exist on the other servers and any deletion event is in ALL of their modlogs. ANY instance can "undelete" or revert the removal, or just ignore the deletion request all together (or roll back the database, or any number of operations to revert a change). The items doesn't just go away. The "origin" doesn't have all that much power to force other listening servers to do anything.
This also extends to comments. I run my own small instance with me and a few friends. My server never had serious downtime because it's just us. Our access to larger instances never "vanished" even as their sites went completely down. The local content is effectively cached regardless of the state of the origin server.
If the host server just straight up ignores turkey then they’ll block all servers that host Mastodon
Good luck with that... There's a lot of servers that can talk the same federation protocol. You're not going to get them all. Forget all the normal means of bypassing blocks... you have so many fediverse and threadiverse servers to attach to in order to access largely similar content.

I wish I could find something like this (low power kinda thing) that could take like 40 sata ssds.
I have a whole stack of 500 GB ssds from a datacenter decommission that I've been sitting on.
The 2TB units found their way into my ceph cluster... but those machines are live vms... A smaller little guy that can stack all these 500 gb would be nice to give to my cousin or something and use as offsite backup.