When it scales multiple orders of magnitude better? Web archives are massive and have non-trivial storage costs.

What would be interesting would just be storing a Merkle tree for archive hashes so many parties could verify that a much smaller number of copies haven’t been modified.