Heinera said:
Then you have the metaphysical loopholes, that cannot even in principle be falsified by experiments. I have to side with Popper on this one: It's not science.
I disagree. Loophole is loophole. One has to close it.
In the case of a metaphysical loophole, it is closed by accepting, as a sort of axiom or fundamental principle, something postulate which prevents it. This postulate, taken alone, cannot be tested by observation.
But this does not make such a postulate unphysical, not even from Popper's point of view. Popper has recognized from the start that not every particular statement of a physical theory can be tested, that one needs the whole theory to get predictions about real experiments. Then, answering Quine's holism, which claims that a single theory is not enough, but the whole of physics is necessary to make experimental predictions, he has recognized even more, namely that often even a whole theory taken alone is not sufficient to derive any nontrivial, falsifiable prediction. Each real experiment depends on a lot of different theories - in particular, theories about the accuracy of all measurement instruments involved.
Just for example that even famous theories taken alone do not give anything, take GR. Whatever the observed distribution of matter, and whatever the gravitational field, by defining dark matter as T_{mn}^{dark} = G_{mn}-T_{mn}^{obs} the Einstein equations of GR can be forced to hold exactly. One needs additional assumptions about properties of dark matter to derive anything from the Einstein equations. Else, all what is predicted by GR is nothing more than what is predicted by all metric theories of gravity - namely that what clocks measure may be described by a metric.
The point what makes it unnecessary to accept Quine's holism is that one can test the several theories involved in each actual experiment in other, independent experiments. This is, in particular, how one solves the problem of theories about measurement devices. You can test the measurement devices in completely different experiments, and this is what is done with real experimental devices. Say, their accuracy can be tested as by comparison with other devices, or (for the most accurate ones) by comparing other devices of the same type.
But, even if we can reject Quine's holism, the other extreme that single principles, taken alone, should be falsifiable, is nonsensical too.
But, once we cannot test them, taken alone, why should we accept them? There are some good reasons for accepting them.
For example, compatibility: Even if we have no TOE, a principle may be compatible with all the best available theories. Another point is what would be the consequence of rejection: It could be that, once it is rejected, one would have to give up doing science, because, if the rejection would be taken seriously, no experiment could tell us anything nontrivial. Superdeterminism would be of this type. Similarly a rejection of Reichenbach's principle of common cause: Once it is rejected, there would be no longer any justification to ask for realistic explanation of observed correlations. The tobacco lobby would be happy, no need to explain correlations of smoking and cancer, astrologers too, because the discussion about astrology would be reduced to statistical facts about correlations - are correlations between positions of planet with various things in our lifes significant or not, and the major point that there is no causal explanation for such influences would disappear.
So, there are possibilities for strong arguments in favour of physical principles, even if they, taken alone, cannot be tested.