Very interesting. Thanks for the detailed testing and results. That should make it much easier for Khaled to reproduce and fix the bug. If I were to make a guess, I'd imagine that the Western encoding is trying to display a character by combining the $chr(3) with the katakana... basically what multibyte is meant to do in the first place, but just getting confused over which bytes to combine. Of course, I may be way off there.
I'll test this out at home later today and see if I can make it work. Unfortunately, I don't have a Japanese locale. I'm not sure if that will matter, but I'll try anyhow. I'll probably run it through a few different versions of mIRC to see if it affects "all" versions (of the ones I test) or if it might be a change made somewhere along the road that caused it. If that turns up anything, it might help to point out exactly what the cause is.
Thanks again for all the testing and detail. That kind of bug report makes fixing bugs much easier!