State synchronization failure during Cluster Agent disconnection

Rancher Manager Cluster Agent

Rancher utilizes a Cluster Agent deployed within managed clusters to maintain communication and state synchronization with the Rancher Manager. This mechanism ensures that configuration changes made in the centralized management plane are propagated to the downstream Kubernetes clusters.

When a managed cluster experiences an extended disconnection from the Rancher Manager during a Kubernetes version upgrade, the synchronization of the cluster state may become inconsistent. This creates uncertainty regarding which configuration updates are applied once connectivity is restored.

Does the Rancher Manager implement a specific reconciliation queue to resolve state conflicts after a prolonged agent outage? How is the sequence of pending configuration changes prioritized during the reconnection phase?