This is a running counter of every IGMP leave message received across all multicast services on the device, the counterpart to the join-message counter. An admin investigating a sudden mass drop in multicast viewers would check whether this counter spiked correspondingly, distinguishing a genuine wave of legitimate leaves from some other cause, like a query timeout aging groups out without explicit leaves. A leave-counter spike that doesn't correspond to any known event, like a scheduled program ending, might warrant investigating whether something triggered mass unintended departures, such as a brief upstream outage.