It is a question of timing.

If you want to research a problem, first you consider alternative hypotheses, then you test them experimentally, and then you throw away the ones which did not pass the experimental test.

At the moment of generating hypotheses, you want diversity. Of course, your time is limited, so you shouldn't bother with hypotheses with huge complexity and epsilon prior probability, so you do some filtering anyway. But you want to have a few competing hypotheses. (You want to keep the hypothesis that the 2-4-6 rule could work for odd numbers too.)

Only later, when some hypotheses are experimentally disproved, you can safely ignore them. (More precisely, there is always a probability that the experiment was wrong. But a good experiment, or many repeated experiments, can move the hypothesis to the epsilon zone.)

So the question is which opinions are unwelcome in universities because they were experimentally disproved, and which opinions are unwelcome, because they are already unacceptable at the hypothesis generating phase.

To which category would the hypothetical supporter of South African Apartheid belong? What exactly are his claims, and which of them have been tested?