The image registry operator and ingress operator use the `/metrics` endpoint for liveness/readiness probes which in the case of the former results in a payload of ~100kb. This at scale can be non-performant and is also not best practice. The teams which own these operators should instead introduce health endpoints if these probes are needed.