Is the broad histogram random walk dynamics correct?