[KirinDave]

I'm happy to be wrong, but I feel like I don't have to overcome the entire network's hashing, I need to overcome several individual instances. If I could isolate and attack individual members of the cluster quickly, I could begin to leverage their role in the cluster as they join, convinced that a single block must be true.

I don't need to overcome the network, I need to invite the network to have arguments with itself, by finding a way to introduce widespread partitioning of the network.

In this, the preference to longer hash chains seems like a good idea in a unified clock model but a somewhat optimistic decision in a split clock world.

[DINKDINK]

>If I could isolate and attack individual members of the cluster quickly, I could begin to leverage their role in the cluster as they join, convinced that a single block must be true.

To do this you would need to have the equivalent hashing power of the network. Peers expect an nominally equivalent amount of Proof of Work for the difficulty adjustment. You would also need to guarantee that the peer connects to _no other_ peers

Your thinking is good adversarial thinking but it's already covered in the protocol.