"Who would make such a decision? And how would you mitigate the damage potential of a "rogue mod"?"

Such decisions are easy to make (by anyone) if the parameters for violations are clearly defined. First of all, rogue mods would be quickly terminated once a pattern of abuse becomes evident, then to mitigate the damage from a rogue mod, compensate the victim in some manner, perhaps at the expense of the offending mod. Secondly, if the bots are programmed properly, there would be almost no need for mods in the first place. Both Google and Facebook have incredibly sophisticated systems for identifying and removing posts which violate their terms of service. i.e. When was the last time you saw porn on Google or Facebook? Rarely, if ever does such posts get past their respective bots. Of course, I realize that few if any platforms, can possibly expect to match the sophisticated systems utilized by Google or Facebook, but if the desire is there, it can be vastly improved over what it is now.