Yes, exactly, you are right.
Obviously "checkpoints" are the solution.
All you have to do is get the network to agree that this "save state" at at this point in time is valid... and then the network remains in consensus.
The thing about @dan is that he's an idealist so it makes sense that he would veto the idea.
Idealism and Reality need to link up somewhere.
Even if one of these checkpoints was corrupt anyone could do a full replay to prove it.
Doesn't seem like a big deal.
Full replay is not sustainable at high transaction rates. In the case of eos snapshots are necessary if you don’t want to take months to replay.
So you serialize all the info in the database... hash it... come to consensus on the hashes... then you can seed one database with another and hash the serialized info to know it's correct? This is how I imagine it working I don't actually know things.