The code that reads the locale files was updated to use InputStreamReader with a forced character set of UTF-8 rather than the standard FileReader method.
The problem with the old method was that files that are read are assumed to be in the encoding of the default character set of the platform.
Sometimes a server will not have a default locale of UTF-8, but rather a language specific locale and thus when the locale file is loaded, the characters are incorrect.