Same happens to me. The regex stops working.
I did some investigation and found:
Looks like the reason for this is an invisible “Zero-width space” (%E2%80%8B) characters get added to the requested URL somehow.
If I copy a URL from 404 log page and then URL-Encode it, I get these characters in the URL: %E2%80%8B
Of course, the regex would not match those characters.