multi-arch tuning operator assumes the implict architecture of an image, and can be wrong
    

    openshift 4.16, 4.17

    only upon a specific assignment of the operator pods to a specific arch

1. Setup a Multi-Arch Compute cluster with arm64/intel, power/intel
2. Cordon the primary compute architecture
3. Install the Multi-Arch Tuning Operator (if control plane is power, this should install on Intel)
4. Create a single arch image which does not declare the architecture. This should be in the control plane architecture. Non-manifest-listed without architecture declaration.
5. Create a Deployment which uses the Pod
6. See the Pod deployed to the wrong architecture.

   the incorrect arch is assigned affinity

     the correct architecture or the control plane architecture is assumed.

    I think this is because of WantedPlatforms where the analysis was executed from an intel machine. https://github.com/containers/image/blob/6692ffcc9970763386a3ac8954f60112fb8f33ae/internal/pkg/platform/platform_matcher.go#L166