can anybody shed some light on the significance of this...
I cannot provide an official documentation, only practical experience.
The use case is that the router with dual-peer policy (typically acting as an initiator) establishes an IKE/IKEv2 SA with both peers (typically acting as responders) but only asks the first one to establish a data SA according to the policy. If it loses the IKE/IKEv2 SA with the first peer, it establishes the data SA to the second one. Once the IKE/IKE2 SA with the first peer re-establishes, it does NOT move the data SA back to the first peer; it only happens if the IKE/IKEv2 SA with the second peer gets lost. If you want the data SA to return to the first peer when it recovers, you have to use a script for that.
If the remote peers are also Mikrotiks, they have to either create the corresponding policy dynamically and have static routes via each other to the subnet reachable using the traffic selector of that policy, or have the policy configured statically but only advertise themselves to the adjacent routers as the gateway to that subnet if the policy is active.
In the first case (dynamically generated policies), the dynamically generated policy overrides any "normal" routes, so if the router where the policy has been generated receives a packet for the remote subnet, it forwards it using the SA; if the policy is not generated, it forwards such a packet to the other responder as the static route says. In the second case, a periodically scheduled script is necessary to translate the activity of the policy into advertising of the subnet via OSPF or BGP or into raising a priority of a VRRP interface. A mix of both approaches is possible.