doc/customizing.txt has a section "8-bit character set support in Web
interface" describing how a character set other than UTF-8 can be used
for sending web pages to the client, given an @charset form variable
which may result in a roundup_charset cookie being set.

The stated reason is "Unfortunately, some older browsers do not work
properly with utf-8-encoded pages (e.g. Netscape Navigator 4 displays
wrong characters in form fields).".  That text was added in 2004 - that
is, those browsers were "older" even then.  I think it would be
reasonable, in 2018, to say that those browsers are now completely
irrelevant and web pages sent to the client should always be sent in
UTF-8, so simplifying the code.

(I noticed this when fixing how actions returning binary data were
handled with Python 3; the re-encoding would break such actions for
Python 2, if anyone ever tried to use them together with the @charset
feature, which probably no-one ever did.)