What happens to in‑flight Kubernetes API calls when an Argo CD sync is cancelled?

When a user initiates a sync cancellation in Argo CD (via CLI or UI), the controller treats the request as a best‑effort signal to stop the ongoing operation. According to the documented behavior, Argo CD does not interrupt Kubernetes API calls that are already in flight; it allows the current step to finish before marking the sync as cancelled. This raises uncertainty about the exact moment the sync status transitions to Cancelled and whether any resources created or modified by the in‑flight calls remain in the cluster.

Additionally, if a sync timeout expires while a cancellation request is pending, it is unclear which condition takes precedence and how the final status is reported.

Does Argo CD provide a way to force immediate termination of active API calls on cancellation? How does the interaction between a pending cancellation and a sync timeout affect the reported sync status and resource state?