Currently Bitcoin has a delay of something like 50 blocks before the block reward is spendable. This imposes a minimum size on the amount of float needed to pull off the attack.

One strategy any proof-of-work chain can use is to just increase the delay. This doesn't mitigate the attack vector though so much as just make it a bit more expensive to get started. It also has imposes some extra costs on miners and may discourage smaller miners from even joining the network.