Value-at-Risk: is lacking in sub-additivity just an annoying technicality?