When you setup cluster peering, you *must* setup a full mesh. Each node on both sides must have an InterCluster LIF. Each and every LIF MUST be able to ping EACH and EVERY LIF.
If you put yourself in a scenario where you have multiple mirrors:
Clus1 -> Clus2
Clus1 -> Clus3
BUT Clus2 CANNOT see or reach Clus3 then you need to make a modification!
In this case, on Clus1, you must create an IPSPACE for one of the two mirrors. Why? If I setup all IP on Clus1, Clus2 can only get to half of them and cluster peering will fail eventually. Same with Clus3 (it can only get to half the IPs).
By creating the IPspace on Clus1 for the Clus1-Clus3 relationship, Clus2 will only ever know about the IPs it communicates with and Clus3 will only know about the IPs it needs to communicate with. You basically isolate the traffic, kinda like a VLAN in so much as it prevents cross-talk between the IPspaces.
the logic is simple. By setting up the full mesh every node can talk to every node. The mirror finds the two nodes involved as a source and a destination and uses those. If the source or destination volume moves so will the LIF being used to compensate for the volume move