If you meet a paperclip maximizer, pulling out your gun could be a moral response. No, it wouldn't mean "might makes right"; the causality goes in the opposite direction: in this specific situation, force could be the best way to achieve a moral outcome. We use violence against e.g. viruses or bacteria all the time.
With humans it's complicated because we actually don't know our own values. What we feel are approximations, or deductions based on potentially wrong premises. So there is a very real possibility that we will do something to maximize our values, only to realize later that we actually acted against our values. Imagine an atheist reflecting on a memory where as a former believer he burned a witch. (What is the strategy he could have followed as a believer, to avoid this outcome?)
So we have some heuristics about moral reasonings that are more likely to change, or less likely to change, and we kinda try to take this into account. It's usually not explicit, because, well, being open about a possibility that your values may change in the future (and debating which ones are most likely to) does not bring you much applause in a community built around those values. But still, our moral judgement of "hurting random people is evil" is much more stable than our moral judgement of "we must optimize for what Lord Jehovah wants". Therefore, we hesitate to torture people in the name of Lord Jehovah, even when, hypothetically, it should be the right thing to do. There are people who don't do this discounting and always do the right thing; we call them fanatics, and we don't like them, although it may be difficult or impossible to explain explicitly why. But in our minds, there is this intuition that we might be wrong about what the right thing is, and that in some things we are more likely to be wrong than in some other things. In some way we are hedging our moral judgements against possible future changes of our values. And it's not some kind of Brownian motion of values; we can feel that some changes are more likely than other changes.
And this is probably the reason why we don't pull a gun on a person living immoral, but not too horrible life. (At some level, we do: like, if there is a terrorist saying that he will execute his hostages, then obviously shoot him if you can.)
I'm not quite sure what your point is and how that relates to what I have written.
The paperclip maximizer, the fanatic and the terrorist all violate agency ethics and the virus is not even an agent.
If you are opposed to my explanations, can you find an example where retribution is justified without the other party violating agency or where someone violates agency while a retribution in kind is unjustified?
Preface
I have trouble expressing myself in such a way that my ideas come out even remotely like they sound in my head. So please apply the principle of charity and try to read how you think I thought of it.
Tit for Tat
Tit for Tat is usually presented in a game between two players where each chooses to either cooperate or defect. The real world game however differs in two important ways.
First, it's not a two player game. We make choices not only on our single instance of interaction but also on observed interactions between other players. Thus the Advanced Tit For Tat not only defects if the other player defected against itself but also if it could observe the other player defecting against any other player that employs a similar enough algorithm.
Second, there is a middle ground between cooperating and defecting, you could stay neutral. Thus you can harm your opponent, help him or do neither. The question of the best strategy in this real life prisoners dilemma is probably still unanswered. If I see my opponent defecting against some of my peers and cooperating with others, what do I choose?
Agency
The reason why there even is a game is because we can deliberate on our action and can take abstract thoughts into account that do not directly pertain to the current situation, which I think is the distinguishing factor of higher animals from lower. This ability is called agency. In order to be an agent a subject must be able to perceive the situation, have a set of possible actions, model the outcomes of these actions, value the outcomes, and then act accordingly.
We could act in such a way that infringes on these abilities in others. If we limit their ability to perceive or model the situation we call this fraud, if we limit their set of possible actions or their ability to choose between them, we call it coercion, if we infringe on their ability to value an outcome, we call it advertising.
Ethics
I propose that the purpose of our moral or ethical intuitions (I use the two words interchangeably, if there is a distinction please let me know) is to tell us whether some player defected, cooperated or stayed neutral, and to tell us who we should consider as having a close enough decision algorithm to ourselves to 'punish' third players for defecting against them. And I further propose that infringing on someones agency is what we consider as defecting.
Value Ethics
Utilitarians tend to see defecting or cooperating as pertaining to creation or destruction of values. (Edit:) Three things bother me about value ethics:
1. Valuations between different people can't really be compared. If we shut up and multiply, we value the lives of everybody exactly the same no matter how they themselves value their own life. If there are chores to be done and one person claims to "not mind too much" while the other claims to "hate it with a passion" we can't tell if the emotional effect on them is really any different or maybe even the other way round.
2. It makes you torture someone to avoid an insanely huge number of dust specs.
3. It makes you push a fat man to his death.
Agency ethics
Instead I propose that defecting in the real world game is all about infringing on someone's agency. Thus we intuit bankers who destroy an insane amount of wealth while not as good people still as neutral because they do not infringe on agency. At least that is my moral intuition.
So infringing on agency would make you a bad person, while not infringing on agency doesn't make you a good person. What makes you a good person is increasing value. Maybe agency is more fundamental and you cannot be a good person if you are a bad person, but maybe you can be both. That would create cognitive dissonance in people who consider ethics to be a singular thing and don't see the distinction, and that might be at the root of some ethics discussions.
Evil
In my version of ethics it counts as evil to push the fat man or to switch the tracks, as that would mean deliberately causing a death of someone who doesn't want to die. I would let the five die and not feel guilty about it, because I am not the cause of their deaths. I make a fundamental distinction between acting and not acting. If I hadn't been there the five would still die, so how could I be responsible for their deaths? I am aware that this view makes me evil in the eye of utilitarians. But I see less people acting consistent with utilitarianism than I see people arguing that way. Then again, this perception is probably heavily biased.
Conclusion
I don't really have a conclusion except of noticing that there exists a disagreement in fundamental morality and to inform you that there exists at least one person who considers infringing on someone's agency as defecting in a prisoner's dilemma.