pub async fn probe_current_cluster_multi_head(&self) -> bool
Re-reads the current cluster’s worker rank addresses from that cluster’s own system properties and records whether multi-head operations are usable there.
This is a capability probe, not a health check: a cluster whose head node is up is usable whatever its workers are doing, and the answer here only decides multi-head versus head-node-only. Callers in the failover path must not treat false as a reason to reject the cluster — doing so kills a failover to a cluster that would have accepted the work through its head node.
The refresh matters because the cluster list is populated once during connection setup and never refreshed per failover; without it the probe would validate a construction-time snapshot of a topology that may have moved since.
Resolves addresses with the connection’s hostname regex, never a caller’s. What this writes — the cluster’s worker addresses and whether multi-head is available there — is connection-scoped state that every multi-head object reads, so letting one object’s filter decide it would let that object’s address selection determine a fact about the cluster for all of them.
Whether multi-head operations are available on the current cluster.