Last time I checked, the Korean wordpress translation database (in translation.wordpress.com), about half of the translated strings were in the UTF8 encoding and the other half were in the EUC-KR encoding.
I suggest the system to choose and force one encoding in the translation database. Contributors (voluntary translators) doesn’t know that there are more than one encodings, so enforcing the encoding must be the burden of the system, or the moderator.
I’ve heard that if you install wordpress by yourself (installing from wordpress.org as opposed having an account in wordpress.com), Korean translations actually work there (it doesn’t work in a wordpress.com blog even if I choose Korean from the blog preference), so another suggestion is that rather than growing the database in translate.wordpress.com from scratch, we should first import from the already established database.