Not sure why I didn't use OP_2DROP in the hashing benchmarks? I think those can be made a little more expensive by saving OP_DROPs, doing lots of OP_2DROPs at the end of the contract, and using those saved opcodes for more hashing.