For a threat to have the desired effect the subject must be convinced the threat is credible. Preparations are essential.
Bgold showed it is easy enough.
Just predispose the "combination strategy" and "combination factor (PoW#1)" ... "combination factor (PoW#n)" fields, and play with them a little. Then the (0, 1, ... , 0) position is exactly as today. Not to be disruptive, we'll try normalized(0, 1, 0.1, ... , 0) next - or (0, 0.97, 0.03, ... , 0), you know, just for the heck of it. Hm, and let's see what happens by combining them differently. Who's playing "bad guy" for this round of simulations? How many ASIC miners do you have?
The parameters could be tweaked by hand at first, just to get a feel for it. Later on, they could auto-adjust depending on measurable blockchain/mempool metrics. I'll let your gaming fantasy get wild here.
No need to freak out, gentlemen. The parameters on main net are still fixed at normalized(0, 1, 0, 0, ... , 0) = (0, 1, 0, 0, ... , 0). That's plain ol'corn for you. This is only a testnet game, of course. Nice though, eh?
That would really make a point.