The NSX for vSphere 6.3.3 controller node has a periodic clean-up task that deletes a status file required by the API server if the API server has sufficiently low activity. After the file is deleted, any new connections from the NSX Manager to the API server will fail.
NSX Manager monitors and updates the controller cluster using REST API calls to the controller-cluster members.
The NSX Manager does maintain a persistent connection to each controller's API server for this purpose. Until the connections are disrupted (physical network issues or restart of NSX Manager) the NSX Manager continues to have access to the controller for cluster monitoring, NSX logical switch & router creation and modifications. Only operations such as Traceflow and Central CLI that do not use the same persistent connection will fail. If external events disrupt the persistent TCP connections, NSX Manager will lose the ability to make API connections to controllers.
Note: The controller API server's role is only for management plane access between the controller and NSX Manager. Disruptions to the API server will not have any impact to the controller-cluster operations or the control plane and dataplane states of NSX. Due to the fault-tolerant design of the distributed controller cluster, NSX Manager continues to be able to update the entire controller cluster as long as it has API connectivity to at least one of the controller nodes.