Current Knative Serving has an issue discussed in SRVKS-547 and we need to wait for upstream's fix. So until upstream fixed it and serverless-operator started using the fixed version, we should document it as a known limitation.

[DRAFT]

• Knative serving has "fixed" CPU minimum request `25m` for queue-proxy. If your cluster have set any conflicting value like defaultRequest for CPU more than 25m, your knative service fails to deploy. As a work-around, you can configure resourcePercentage in each knative service.

Example:

spec:
  template:
    metadata:
      annotations:
        queue.sidecar.serving.knative.dev/resourcePercentage: "10"