This is a great point about handling printable vs non-printable characters that I originally missed when I read wc code. Thank you for pointing this out!
I ask because the essay and your comments until now show no insight from reading the code.
I find it difficult to understand how anyone could miss that (rather significant) part of the core algorithm, and then assert the differences are due to only "modulo intended Unicode space handling" and the like.
Until now I had assumed you had a lay understanding of wc, and had not read the code.
On a meta level, I wasn't even considering that the notion of a word might be different from "a sequence of characters that aren't space characters".
Live and learn indeed.