Carl, if putting two drugs together on the same shelf is a bad idea, it's bad because doing so is fraudulently deceptive, and so can be handled simply by policing fraud.
Indiscriminate use of antibiotics isn't quite "germ warfare", but I agree it involves a negative externality that can qualify as aggression. The flip side of that coin is encouraging contagion by not fighting germs hard enough. Prevention of contagion and conflagration and floods, by addressing the "holdout" market failure, is one of the very few forms of pure public goods (like national defense) that in my judgment merit government intervention. For details, see the handy table at http://libertarianmajority.net/public-and-private-goods.
Susan is correct to criticize you for suggesting "net benefit" as the way to evaluate any government intervention. I wouldn't even agree that a government intervention is justified merely because it is Kaldor-Hicks efficient. However, her anarchistic criterion of "voluntariness" is inadequate too. The way for a libertarian to judge a government intervention is to ask whether it is both 1) Kaldor-Hicks efficient and 2) effective at protecting life, liberty, and property. This combines both your idea and Susan's, while avoiding the Scylla of nanny-statism and the Charybdis of anarchism.
Alas, concepts like Kaldor-Hicks efficiency and pure public goods require about 100 words to explain, and so they are much less politically viable than concepts like "provide any net benefit" or "never initiate force". The real trick in politics is to design political institutions and processes that 1) implement the political principles that the local primates would want if they understood them while 2) not disenfranchising them.
