"the test is particularly appealing to people with a certain personality (i.e. "hackers"), and that the test happens to label these people identically (i.e. "NT")."
Isn't this the only claim that proponents of the test make? It's the whole point: people with different personalities are supposed to show up differently on the test. It's not clear to me why you seem to think that the claims for the test are actually claims against it. :)
Well, the Meyers-Briggs test claims to be able to discriminate many different types, not just NT variants. But NT seems to be a type that is particularly well-identified by the test. You don't often hear SJs (for example) raving about the accuracy of their type classification.
Simpler example: if I created a test that appealed primarily to cooks, written using language and questions that related to food and gastronomy, you probably wouldn't be surprised if that test "profiled" the characteristics of a large number of professional chefs. You could argue that such a test is therefore a good discriminator of professional chefs, but that doesn't necessarily make it an interesting assessment of personality.
Point is, maybe there's something about the analytical nature of the people described as "NT" that leads them to find this test appealing/accurate/relevant. You can control for this phenomenon in an experimental setting, but you can't point to the results here, and suggest that it validates the test because the results are exceptionally skewed.
Isn't this the only claim that proponents of the test make? It's the whole point: people with different personalities are supposed to show up differently on the test. It's not clear to me why you seem to think that the claims for the test are actually claims against it. :)