In the kube-prometheus-stack, for the burn rate related queries, I am having some difficulty understanding the query for detecting api server requests that are too slow. For example, here

My understanding is that the three buckets le=1, le=5, and le=30 will contain some of the same elements, and I do not understand how taking this away from apiserver_request_duration_seconds_count wil leave you with queries that are too slow