See http://www.unicode.org/reports/tr10/#Searching When doing a case insensitive search, comparing strings with foldCase is not enough to get the same results as ICU. ICU uses the primary level strength when the search mode is not case sensitive, this means that depending on the current locale some accented characters match to its not accented version. For example in Spanish "á" matches against "a", in German "ß" matches against "ss", in Danish "aa" matches "å", etc. This makes some tests like fast/encoding/find-russian fail.
GLib backend is gone per bug 109627, closing.