Currently, the gather-extra step snapshots a single prometheus-k8s-* pod's data. Depending on which Prom was up and scraping when, that can leave holes in the gathered assets, such as:

alert KubePodNotReady fired for 60 seconds with labels: {namespace="openshift-oauth-apiserver", pod="apiserver-567f9996b4-t5r86", severity="warning"}

in this job which is completely absent from the gathered Prometheus dump. Is there a recommended procedure for gathering a Thanos-aggregated dump with data from both Prom instances?

Might be related to API-1146?