But if you can't tell, then it is credible as a threat.
You could have dropped half that comment and just put this as the first sentence. And I find it funny how much you use the word "clearly" when we're talking greyscale.
"If I knew where you lived, you'd be toast!"
...doesn't mean that they won't try to *find out* where you live. So whoop--it could happen! Therefore that one's credible.
Considering how notoriously hard it is to tell on the Internet whether someone is being serious, I would only exclude those threats that are physically impossible (your Cardassian example). In which case it's possible to end up with 90% of threats received being "credible." Maybe not likely, though? It is Internet trolls we're talking about here, but SWATing is a thing, too.
You can get the whole thing from the semantics of the words "credible" and "non-credible," by checking if it is non-credible. If it has something as mentioned above that makes it "non-credible," then it is not credible.
I'm not finding extracting an objective definition from this circular definition as "clearly" easy as you claim. Saying A = everything !B doesn't work when A and B overlap, because you can argue that B = everything !A and now we have an incompatible center of the Venn diagram which is both A, therefore !B, and B, therefore !A, which was kind of my point (in retrospect ;)