How this principle related to Golden Rule and Eye for an Eye is called?

Question

Had long admiration to this example:

When criminals were sent to isolated island centuries ago they suddenly realized that it's not fun anymore when everyone is criminal. And switched to one of most friendly and lawful behaviours compared to other World (Australia intended).
Or same for littering: people who have zero littering policy won't even notice any troubles when they put together. However, people who do litter, when they concentrated together, will very often try to escape this place for better life - they need to switch behaviour model.
Focus in both cases is on switching behaviour.

This can be propagated on the countries level:
For example, a principle of protecting people who talk same language or have same religion in other countries will produce infinite conflicts, at least due to migration flows. However, for example, if Afghanistan will offer discriminated Muslims opportunity to move to Afghanistan - this works fluently and will not create major wars (propagated as imperative for all countries behaviour).

The best name I found is What if everybody else did this test.

It also looks great to propagate this to best punishment strategy: if someone violates this test, the sentence should be to ensure this person gets a reflected back "What if everybody else did this" experience.

1. Please provide a practical example where any of these concepts (rule or punishment) fail?
   Obviously, the assumption is that participants are relatively similar OR that difference will get mirrored with years, i.e. a strong participant will be weak under similar situation years later.
   _{This is not bad assumption because now it's close to proven all humans originate from same tribe 2M years ago. And migration started much later around 200K years ago. Both incomparably smaller than evolution time (dolphins and humans common ancestor is 60M+ years ago, and we still incredibly similar on embryo and behaviour).}
   Considering possibly non-optimal result fail - there is an idea that main law system goal is to decrease crime levels, and raw justice is only second priority. But if recall is correct, there were some studies which suggest that making less harsh punishments than Eye for an Eye _{(modern version probably includes psychological damage to people in the area too)} does not principally influence repeat chances.

2. How this combined approach test+sentence guide is called? Best I can think of is rule of thumb for behaviour for civilized people/countries. But that's too long and too casual.
   Considering the Categorical Imperative name: the main concern is that if to search for CI, there is practically no mention for what-if test anywhere.
   Another concern is perfectly naive and intuitive (touched philosophy years ago and updated a bit now), but my intuition tells that CI focus is personal actions and What-if-test focus is environment between people.

Answer 1

1. it looks to me both principles are axiomatically correct.

Rule: it is another way to say don't do anything with negative total effect in the future. Only problem is to count the total effect correctly - people are considered fundamentally same (evolutionary origins reason elaborated in the question above), and total sum counted over many years, and so on.

Punishment: it basically ensures total zero effect for the actor who is doing harm making his action pointless.

In both cases, an isolated island (or closed city) thought experiment just helps a lot to properly evaluate the effect of harmful action. If someone steals, the total effect is incomparably higher than cost of item. Total effect is probably the difference between a place when nobody steals and a place where everyone steals.

2. ?

Actually… If everyone starts to answer their own questions on Stack Philosophy, what is total effect?