What does this error mean: The last watch has yet to terminate as expected, will force start another watch.

[robvanmaris]

I'm seeing the error message below in our logs, I'm not sure what it means. Can someone explain?
The last watch has yet to terminate as expected, will force start another watch. Please report this to the Fabric8 Kubernetes Client development team.

Context: we use KubernetesClient to monitor the state of the pods in our statefulset. After running for months this is the first time we see this error message, two times on two different pods, 140 seconds apart.

[shawkins]

The logic is hesitant about starting additional watches because there can be resource issues with too many concurrent watches. Under normal circumstances, the closing of a websocket is bi-directional. On side sends a close, then the other sends back a close. The kuberentes client logic here is expecting that the server will have acknowledged the close, but it hasn't. It's possible this could happen for a benign reason, such as the api server becomes unreachable after the close attempt. If this keeps happening we'll need to see more of a log to determine the full sequence of events. For now if the watches recover, then there's no need to worry.