Based on best practices, if an SSID is configured for a primary and secondary gateway...

With primary and secondary gateway clusters defined on an SSID, APs follow a connection preference list: they attempt to form tunnels to primary gateways first; if all nodes are unreachable, they individually fail over to the secondary cluster.

Preemption governs return behavior (moving APs back to primary when it becomes available), but the decision to fail over when the primary is down is made per AP based on its ability to reach any node in the preferred cluster.

A cluster leader does not command APs to move when the entire primary cluster is down; the APs detect loss of keepalives and independently establish tunnels to the next cluster in their SSID profile.

References (HPE Aruba official materials): AOS-10 Gateway clustering and AP tunnel behavior—AP connection order (primary→secondary), failover decision per AP, and preemption controlling failback.