For a specific cluster Datadog is reporting a change of Max, Min and Current Replicas about every 20 mins for the metric kubernetes_state.hpa.max_replicas . The actual number of pods/containers does not look to be changing neither the config files for the cluster.
Rare cases – This could be a Datadog visualization thing. Using max by rather than sum by looks to have produced desired visualization . Just by getting into incognito mode | private browsing | clearing cache | after few mins | logout and logback in , issue should have been resolved