Its probably related to that bug somehow. One thing that appears to be happening is a gratuitous utf8::encode on the way through mysql.so so that if I have one database field that is riddled with wide-character representations, it shows up in the database ok but another database field that is already decoded, it mangles as though it performed a utf8::encode on a string that was already encoded. The string that is mangled does have utf8::is_utf8 == Test::utf8::is_flagged_utf8 == 1 and Test::utf8::is_sane_utf8 == 1;