Connection Error in Application Controller target cluster communication

The Argo CD Application Controller maintains a reconciliation loop to align the live state of Kubernetes resources with the desired state defined in Git. This process relies on consistent connectivity to the target cluster to perform health assessments and calculate sync status.

When the controller cannot establish a stable connection to the target cluster, the Application resource reflects a Connection Error status. In large-scale deployments, it is unclear how the controller prioritizes reconnection attempts when multiple target clusters experience simultaneous network partitions or API server throttling.

• Does the Application Controller implement an exponential backoff strategy for clusters in a Connection Error state?
• How does the reconciliation loop handle resource health updates during the transition from a Connection Error back to a healthy state?