HyperShift is considering allowing hosted clusters to avoid having a monitoring stack (ADR-30, OBSDA-242). Platform metrics for the hosted clusters contain the data we've used so far for conditional update risks, and those will be scraped by Prometheus on the management cluster, retained for >=24h, and remote-written to a Red Hat Observatorium cluster. Current conditional update risks have never gone beyond max_over_time(...[1h]) to smear over pod restarts and such, so local Prometheus/Thanos should have plenty of data for us.

This ticket is about adding knobs to the CVO so HyperShift can point it at that management-cluster Thanos.

Follow-up work will teach HyperShift to tune those knobs when creating the CVO deployment.