Here are the correct bytes:
$ echo 民國|hd
00000000  e6 b0 91 e5 9c 8b 0a

Yes, with plain g, the article looks great in the *Article* buffer, but
still one big mess in the *Summary* buffer.

In .overview, the line is UTF-8, but there are other non-UTF-8 lines in
.overview, probably causing the whole file to be assumed non-UTF-8 by
emacs as we discussed earlier.
https://www.jidanni.org/comp/configuration/