Was previously not handled at all, therefore displaying request failures
in the test log output. While that was mostly a red herring and
shouldn't have caused any actual *test* failures, it is still better to
handle this explicitly in a realistic manner.