kernelkit
Odd bridge port fastleave behavior
Currently an IGMP/MLD device that sends "leave" will cause the bridge to drop the multicast traffic immediately from the port, regardless of the port's fastleave setting.
The expected default behavior (when fastleave is disabled) is to wait one QRI (query response interval), 10 sec., before removing the port from the MDB filter.
My current guess is that the bridge tracks join/leave for all hosts and groups on a port (IGMPv3/MLDv2) and then decides to just remove the group from the port when there are no more subscribers. I have not verified if the behavior is different when the port is in IGMPv2/MLDv1, or if the bridge sends a group specific query or not, but in my opinion it is not optional to not wait QRI unless fastleave is set (that's the option).
Since the default for the bridge is to flood unknown multicast, I guess one reason for my finding could be that I only tested with one receiver for a group. If flooding was disabled and the last subscriber was instantly removed, that would be a bigger problem.