Summarize pods not running, by Namespace and Reason - I'm having trouble finding the reason

Question

I am trying to emulate this script which summarizes statuses of pods that are not running by namespace and reason.

kubectl get po -A --no-headers |
awk '
  BEGIN {
    SUBSEP=" "
    format = "%-20s %20s  %5s\n"
    printf format, "NAMESPACE", "STATUS", "COUNT"
  }
  !/Running/ {a[$1,$4]++}
  END {
    for (i in a) {split(i,t); printf format, t[1],t[2],a[i]}
  }
' | sort

This script produces output similar to this:

$ notrunning

NAMESPACE                                                  STATUS  COUNT
namespace-01                                     InvalidImageName      2
namespace-02                                     InvalidImageName      1
namespace-02                                Init:ImagePullBackOff      1
namespace-03                                     CrashLoopBackOff      2
namespace-03                                     InvalidImageName      9
namespace-04                                    Init:ErrImagePull      1

I can't find where kubectl is getting the status or reason. I'm trying code similar to this (leaving out some error checking for brevity). I am not getting the results I expect.

type PodSummary struct {
    NotRunning int
    Summary    map[PodKey]int // Map of namespace+state to count
}

type PodKey struct {
    Namespace string
    Status    string
}

func getPodSummary(kubeconfig, cluster string) PodSummary {
    clientset, _ := getClientsetForContext(kubeconfig, cluster)
    pods, _ := clientset.CoreV1().Pods("").List(context.TODO(), metav1.ListOptions{})
    summary := PodSummary{Summary: make(map[PodKey]int)}

    for _, pod := range pods.Items {
        if pod.Status.Phase != "Running" && pod.Status.Phase != "Succeeded" {  // need to check "Completed" also?
            podNS   := pod.Namespace
            summary.NotRunning++
            var pk PodKey
            for _,containerStatus := range pod.Status.ContainerStatuses {
                if containerStatus.State.Waiting != nil {
                    pk = PodKey {podNS, string(containerStatus.State.Waiting.Reason)}
                    break
                } else {  // cannot find it, use this instead.
                    pk = PodKey {podNS, string(pod.Status.Phase)}
                }
            }
            summary.Summary[pk]++
        }
    }
    return summary
}

I was expecting to get detailed reasons why the pods are failing. Instead I got results like "Pending" which isn't helpful or what I wanted.

Here is an example of a pod in a bad state, yet .status.phase says it is Running. How can I do a better test? And what if the pod has multiple containers? How do I choose which to report on?

$ k get po
NAME                     READY   STATUS             RESTARTS           AGE
notreallyrunningpod      0/1     CrashLoopBackOff   5152 (85s ago)     10d

$ jq -r '.status.phase' < notreallyrunningpod.json 
Running

$ jq -r '.status.containerStatuses | .[] | .state.waiting.reason ' < notreallyrunningpod.json 
CrashLoopBackOff