Jiang Xin [off-list ref] writes:

If we want to print "Done" in porcelain mode when there are no
errors. (In dry run mode, ref-update-errors should not be
considered as errors, but the opposite should be regarded as errors).

Hmph, should we allow dry-run to deliberately behave differently
from the for-real execution?  Wouldn't it discard the primary value
of dry-run, to see if it is likely that an operation will succeed
(or fail)?