> If you know of any good UTF-8 corpora for common use cases, I'd be happy to benchmark them.

I think a reasonable approach is the first N bytes of wikipedia for various interesting languages which stress the UTF-8 continuum, e.g., English, French, maybe something in Cyrillic, Greek, Korean, Japanese, Mandarin.

Those are in html IIRC so you should have a good mix of ASCII plus UTF-8.