If an L1 enforces censorship after a large hack, then why not go all the way to implement in-protocol mechanisms to enforce social consensus so that any consensus hack can be reverted?

In the end-state a chain is either a neutral base layer or a social consensus engine.