When there are many resources, the resources panel takes a very long time to update.
Number keep updating and when taking too long they drop back down and increase again.
This is probably since the information is gather from ETCD.
If it would be updated based on metrics, this will load very fast and present the correct value.

The attached screenshot video shows the pods that are in error increasing. 
The value presented in the UI is ~6k, but the actual pods in error was around 12k.
This can be significantly improved by using the metrics like
for pods :

• Total : sum by (phase)(kube_pod_status_phase)  
• Filter by a specific phase: sum (kube_pod_status_phase{phase="Failed"})