Here’s a summary: https://ploum.net/2023-06-23-how-to-kill-decentralised-networks.html

That’s besides the point. Of course it’s always possible to create new communities on new instances, and import posts from various sources, but the original community would be still gone.

If an instance is shut down or becomes unusable for a long time there is no way to automatically migrate users to a new instance. Additionally, there is also no guarantee that all users will move to the same alternative instance. This can also cause unnecessary conflict around which alternative instance becomes the “legitimate” successor.

What you are describing is just a local cache of !lemmy@lemmy.ml on your instance and it works only if it has been populated before the downtime of lemmy.ml. If lemmy.ml never comes back to life nobody can post to !lemmy@lemmy.ml proper. All the communities on in would be dead.

This is not a communication problem. Communities are indeed centralized and if an instance is shut down permanently or loses its data all the communities are gone. This is a big design problem of Lemmy.

Edit: it’s sometimes possible to rebuild new communities on another instance and recover past messages that have been replicated on other instances (if there were full replicas) but this requires all users and moderators to agree on where to migrate and avoid splits and so on.

Indeed. If a big instance like lemmy.ml was to be shut down all the communities would be lost. This is simply not sustainable. Why would users put effort building a community if it could be gone at any time?