I'm getting strange inconsistencies in how some unicode glyphs are rendered in mIRC.
Refer to this comparative example:

Two sets of two lines each. The upper set and lower set should look exactly the same; they're the same exact characters, being read in by mIRC from the same file. As you can see, however, the lower examples use some different glyphs and have anti-aliasing/font-smoothing applied. This inconsistency occurs despite no changes in configuration on my system whatsoever, or even rebooting (in some cases). Unfortunately, I have no way of reproducing this behavior on command.
If it matters, the font I use is Fixedsys Excelsior 3.01, and I'm running Windows XP.