Prometheus query used for pod metrics is:

*container_cpu_load_average_10s{pod="

{#NAME}", namespace="{#NAMESPACE}"}*

but this retrieve with the value set to "sum" a value that is doubled the real one because it read also the parent cgroup value

In my opinion should be better to use:

*container_cpu_load_average_10s{container!="",pod="{#NAME}

", namespace="{#NAMESPACE}"}*

the issue is the same for the other 2 metrics of pods