How should an engineer optimize multicast packet distribution throughout the overlay network in SD-WAN?

This choice describes removing replication from the WAN Edge connected to the source, which is the same concept as D but uses imprecise terminology - the key optimization target is the 'ingress router,' not just any WAN Edge.

Building a shared tree where a WAN Edge learns the multicast source IP describes traditional PIM RP shared-tree behavior, not the SD-WAN overlay optimization mechanism.

Building a shared tree where a last-hop router learns the source IP describes PIM SPT switchover, which is a native IP multicast concept unrelated to SD-WAN overlay multicast optimization.

In Cisco SD-WAN multicast, the vSmart controller distributes receiver membership information so the ingress WAN Edge connected to the multicast source no longer blindly replicates to all WAN Edges; instead it sends traffic only on tunnels with active receivers, eliminating unnecessary replication at the source side.