It's the reason expertise fields end up creating new terms with time. The old terms discriminatory sharpness decrease with time as they get more popular and used in more and more places.

I guess the end game for a language would be when you can no longer tell apart yes and no even when taking into account the sourounding context.

I wonder it there are people who do this kind of linguistic research using information theory, doing a quantified tracking of the collapse of expressivity of terms and more generally languages.