The W3C Link Checker, when I entered <http://cold32.com>, allowed 10 levels of recursion (more than needed), and set it to send the Referer, reported a fault but it was caused by the link checker's own fault. A link in http://cold32.com/image-credit.htm for https://en.wikipedia.org/wiki/Yup%27ik_clothing#/media/File:Eskimos_woman_and_girl_ice_fishing.jpg was interpreted by W3C as https://en.wikipedia.org/wiki/Yup'ik_clothing#/media/File:Eskimos_woman_and_girl_ice_fishing.jpg (interpreting "%27" as "'") but W3C said that was redirected to https://en.wikipedia.org/wiki/Yup%27ik_clothing#/media/File:Eskimos_woman_and_girl_ice_fishing.jpg (interpreting "'" as "%27"). If the Link Checker had not percent-decoded one character, no error would have been reported, because the URL on my page and the destination URL were identical but W3C's was not.

For this bug report, I guessed the component and the version; the version is actually 4.81.