[Just found this. Not posted due to "Bad gateway" at the time. Also, I don't think my feeds pops this blog anymore.]

To be fair to Haldane, his response excludes possibility 3). But he is entirely correct. The problem is that most scientists, and biologists in particular, thinks of their theories in a non-rigorous way as regards testing. That is, evolution is defined to be inclusive and you can add (and exclude) or modify mechanisms and constraints at will.

But when you test a theory you have to freeze its definition (mechanisms and constraints) and test that theory. So Haldane would have to reject the evolution theory where the standard phylogeny predicted no rabbit species in the Cambrian. That doesn't mean such an observation would prohibit all forms of evolution. But it would put a serious constraint on evolution indeed, so much so that very little would remain in the new theory.

The star example (I prefer to think of it as the first that aspired to immortality, Gilgamesh*) is actually putting two theories against each other. Cosmology and star dating. (No, not the Hollywood kind.) I would assume Gilgamesh was dated by looking at metallicity trends.

[*Now I feel like I'm trolling, especially since my feeds always pops this blog, but this is what I like. The greek-egyptian-phoenician abrahamic synchretic myths seems to derive from Alexandria, and so are way younger than the persian mythology.]

When such an age run up against cosmology, it is far simpler to assume the theory (or model, if you wish) that dates stars is in error. And the simpler alternative is apriori the most likely.

One has to remember that, contrary to some claims in the comments, the inflationary standard cosmology allows for correct dating of the universe for the first time. Before its discovery, such ages diverged by a factor of 100 %! After Planck the age error is down to ~ 0.3 % @ 3 sigma (IIRC), nearly as good as the latest dating of the solar system at ~ 0.1 % @ 3 sigma (IIRC).

So we know the age of the universe within /- 30 million years, but apparently the age of some of the oldest stars within +/- 800 million years. Star dating is not only the theory most likely to be erroneous, it has an inherent error of ~ 20 times that of cosmology.