"hg log --style=changelog" has a couple of features I miss in "darcs
changes", listed below.  The first I care about, the other is listed
more incidentally.

If a sequence of patches have the same author and day (date), they are
listed under a single author/date timestamp.  For example,

  Mon Jun 16 14:11:21 EST 2008  Trent W. Buck <trentbuck@gmail.com>
    * Add clarifications from meeting.

  Mon Jun 16 12:28:09 EST 2008  Trent W. Buck <trentbuck@gmail.com>
    * Generate ODF; use more quotes.

  Mon Jun 16 12:27:51 EST 2008  Trent W. Buck <trentbuck@gmail.com>
    * Simple gmake interface to Build.sh and Clean.sh.

simplifies to

  2008-06-16  Trent W. Buck  <trentbuck@gmail.com>

    * Add clarifications from meeting.

    * Generate ODF; use more quotes.

    * Simple gmake interface to Build.sh and Clean.sh.

Emacs' change-log-mode marks up this date/user header correctly,
unlike the current darcs changes output.

Secondly, it automatically adds "changed file" leaders.  For example,
instead of

  * Generate ODF; use more quotes.

it prints (approximately)

  * Build.sh, Clean.sh: Generate ODF; use more quotes.

This latter feature is FAR LESS important to me than the former,
because "darcs changes --summary" is more informative and more
readable in a generic pager.  However, Emacs' change-log-mode
understands the file links in the "hg log --style=changelog", but not
in "darcs changes --summary".

Another feature that might be nice is to automatically list, in
standard notation, the changed procedures / sections (detected by a
heuristic similar to GNU diff -p).  That is, instead of

  * Simple gmake interface to Build.sh and Clean.sh.

it might print

  * GNUmakefile (all, clean, distclean): Simple gmake interface to
    Build.sh and Clean.sh.

this might be a bad idea, because the heuristic is bound to get things
wrong.  In any case AFAIK the point is moot with the current patch
format, since it doesn't contain the right information.

Finally, I anticipate that people will complain their scripts parse
the output of "darcs changes" and they don't want to update them to
accommodate "gratuitous" changes.  My response is the hard line that
those scripts should probably be parsing the output of --xml, since
the default format is for *humans*.