When working on my current project, Science Gothic, early on I played with having the weight, width and contrast axes for the notdef. It looked cool! But then I realized that my notdef would not stand out as much, and that the whole point of the notdef was to look out of place and call attention to itself. That way one sees it when proofing or examining a document.
So I reverted to a generic notdef.
Then recently one of my colleagues working on the project, not knowing this history from early on, did the same thing again, and made a variable notdef that has its weight and width axes. He is less persuaded by the argument against it.
I am curious what others think. Should the notdef be highly generic so it stands out, regardless of font style? Or should it be font-specific and blend in?