Thanks! That is a difficult question and I don’t think it has a simple answer.

I don’t think it can completely preclude any action against the non-aggression principle, because difficult and complex scenarios do ultimately happen, and it is not reasonable to expect people to be completely paralysed to act. Really, it depends on the level of aggression – the less reversible the aggression is, the less easy a mistake is to mitigate, and therefore the more convoluted the situation must be to justify it.

For example, breaking someone’s finger in order to close a door to prevent the spread of a fire seems fairly reasonable. They weren’t deliberately endangering people, but “aggression” is justified in order to save lives. Even if the number of lives at risk is uncertain, it would be hard to argue that breaking that person’s finger was not reasonable.

Killing someone, or causing severe physical or psychological trauma are the real issues here, as they are basically irreversibe. Even in a situation where the Non-Aggression Principle would not usually apply, such as self-defence, this concern about uncertainty is still valid.

For example, if a knife wielding maniac comes into the room threatening everyone’s life, it may still be worth trying not to use lethal force, just in case you are hallucinating and end up harming an innocent person. Alternatively, you could reduce the risk of it being a hallucination by getting confirmation of the situation from people around you (requiring either them to be a part of your hallucination too, or for some sort of mass delusion to be occurring, both of which are even lower probability situations). This additional assurance could potentially allow greater force to be justified.