Running precaching for an upgrade of spoke cluster from 4.10.33 to 4.11.10 the number of images in the crio cache is significantly higher than expected.

Number of images in crio image cache on the spoke node:
prior to precaching: 113
after precaching: 304

The 191 images pulled by precaching are:
170 are quay.io/openshift-release-dev/ocp-v4.0-art-dev with no tag (just hashes)
21 are related to specific operators (kube-rbac, local storage, ptp, sriov, talm, loggin, cloud-event-proxy, and operator index)

I have attached the crio image list pre and post pre-caching.

An additional set of content which is pulled (but not needed) is Operator images for subscriptions which have complianceType set to "mustnothave". When upgrading from 4.10 to 4.11 operators we must include a "mustnothave" policy on the PerformanceAddonOperator subscription (so it gets removed). The pre-cache logic includes PAO in the set of cached content. Output from pre-cache when PAO subscription was set to "mustnothave":
    spec:
      operatorsIndexes:
      - registry.redhat.io/redhat/redhat-operator-index:v4.11
      operatorsPackagesAndChannels:
      - sriov-network-operator:stable
      - ptp-operator:stable
      - performance-addon-operator:4.10
      - cluster-logging:stable
      - local-storage-operator:stable

TALM 4.10.0-202207221207

Always

1. From 4.10.33 cluster, run 'crictl images' to see initial set
2. Create upgrade policies (clusterversion), and common config update to point operator index to 4.11
3. Create CGU, enable=false, precache=true
4. Wait for it to complete precaching
5. run 'crictl images' to see final set

# wc -l all_images_p*
  304 all_images_post
  113 all_images_pre

(Confirmed that all the differences are new images in the "post" set)

precaching to pull approximately the same number of images as exist on the system after initial install.