I do think it is fishy that editor don't update/change when the filesystem provider goes (and returns). Adding @bpasero what the semantics are supposed to be. I think unregistering a file system provider is similar to cases where files are deleted on disk, when a new provider for the same scheme return the editor should probably update.

For readonly-ness we also have #73122 which is about readonly-ness on the level of a file

Originally posted by @jrieken in #109527 (comment)