Noam Yorav-Raphael, I tried to add you because this is pushing back against your patch in issue 8408.  It's been some years now, and nobody has cared enough to pursue it, so I'll just close this if you still don't care ;-)

As doctest's original author, I appreciate why 8408 was done, but don't think it was "a good" solution.  In fact doctest can have no idea whether an example was _intended_ to be run with or without a custom shell's displayhook function invoked to massage the output first.  So more sensible would have been to add a new doctest directive + optional argument, to make the intent explicit.

Which is certainly more work, and hasn't actually come up as "a problem" yet.