vi international character display

["Ian D. Leroux" <idleroux%fastmail.fm@localhost>]

Sometime this year, vi started displaying non-ascii characters as
hex escapes even in a unicode terminal.  I first noticed the change in
the late summer.  I was busy at the time and didn't investigate or
document the change as well as I should have. Here's what I can say
about it with reasonable confidence:
- Before the change, common West European accented characters like é,
  è, ø, ç in utf-8-encoded files displayed as themselves in vi in an
  rxvt-unicode terminal. Since the change they display as hex escapes.
  The change does not affect all characters; I've found, for instance,
  that ? (w with an acute accent) still displays as itself even in
  current vi.
- The environment, which I'm fairly certain I didn't modify around the
  time I observed the change, includes
  EDITOR=vi
  LC_CTYPE=en_US.UTF-8
  VISUAL=vi
  LC_COLLATE=C
  TERM=rxvt-unicode
  COLORTERM=rxvt   # I've also tried setting this to rxvt-unicode
  LC_NUMERIC=C
- It's really a unicode-capable terminal, because those same accented
  characters display as themselves in less or ksh.
- The change happened between February and July of this year.  The only
  commit to vi I see in that window is tnozaki's fix of PR/44455 on
  March 21st.

This is all on amd64, with current updated as of July 28th.

Has anyone else noticed this?  Was the change deliberate (e.g.
to avoid actually broken behaviour on other platforms)?  What other
information can I dig up to turn this into a useful PR?

-- IDL