During MCO updating a node, in certain conditions (e.g., unstable network), a failed cordon request can leave the Unschedulable state of the node instance held by MCO out of sync with the node’s actual Unschedulable state in the cluster, causing MCO to hang on that node until it eventually times out.

    1. Create a MachineConfig that triggers a node drain and apply it.
    2. When a specific node starts updating, intercept the first cordon patch/update request and force it to fail.
    3. Observe the MCP status and wait for the operation to hang and eventually time out.     

    MCO gets stuck on a specific node during update, repeatedly attempting cordon while the node state does not progress

    Even if the first cordon attempt hits a network error, retries should be recoverable, converging to the correct Unschedulable state