We know that fetch_all_by_Transcript() (on a TranslationAdaptor) will
  return a list of translations, where the first one will always be the
  canonical translation (if there is one).  Make use of this fact to
  make the routine slightly quicker.