For clusters with no instances, there is no point in computing either
the score or in running the algorithm. In this case, we exit prematurely
and when running in one-line mode we show dummy information.