I thought your version was "weirdly dependent on context." It was.
So what? I never said it wasn't.
I had tested your original, uncorrected version of length_in_grapheme_clusters() in my Perl script and got results I didn't understand when I called the function in list context instead of scalar context.
Again, so what? You said far more than that, and I responded to the rest of what you said.
Now it's not because you changed it to be more like the one you disparaged earlier.
No, I took no steps closer to reintroducing the silly loop.