Time-consistency of risk measures: how strong is such a property?