When Image Mode is disabled by removing the MachinOSConfig resource, the MCP needs to be updated with the OCP default osImages and when it happens the MCP is reporting a wrong number of ready nodes

IPI on AWS version 4.20.0-0.nightly-2025-06-27-040027

Always

1. Create a MOSC to enable Image Mode in the worker pool
2. Wait until the new image is create and applied to the nodes


$ for node in $(oc get node -l node-role.kubernetes.io/worker -o name); do oc debug -q $node -- chroot /host rpm-ostree status; done
State: idle
Deployments:
* ostree-unverified-registry:quay.io/mco/layering@sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                   Digest: sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                  Version: 9.6.20250626-0 (2025-06-27T11:35:39Z)
State: idle
Deployments:
* ostree-unverified-registry:quay.io/mco/layering@sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                   Digest: sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                  Version: 9.6.20250626-0 (2025-06-27T11:35:39Z)

3. Delete the machineosconfig to disable Image Mode


$ oc delete machineosconfig 
machineosconfig.machineconfiguration.openshift.io "mosc-worker" deleted


4. Check the numbers that the worker MCP is reporting

$ oc get mcp
NAME     CONFIG                                             UPDATED   UPDATING   DEGRADED   MACHINECOUNT   READYMACHINECOUNT   UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
master   rendered-master-f9398fc9986d7605b20e103c9bdc56a0   True      False      False      3              3                   3                     0                      4h10m
worker   rendered-worker-55c9b4e5fab5c47c17d76fc81f9170d7   False     True       False      2              1                   1                     0                      4h10m

The worker pool is immediately reporting 1 updated and ready  node when it is not true. Both nodes aren't still updated or ready

$ for node in $(oc get node -l node-role.kubernetes.io/worker -o name); do oc debug -q $node -- chroot /host rpm-ostree status; done
State: idle
Deployments:
* ostree-unverified-registry:quay.io/mco/layering@sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                   Digest: sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                  Version: 9.6.20250626-0 (2025-06-27T11:35:39Z)
State: idle
Deployments:
* ostree-unverified-registry:quay.io/mco/layering@sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                   Digest: sha256:17c67e0aa08f43cd6e7e1f5d0b1d437e034aa411dcf95b2c3a01a79f64686399
                  Version: 9.6.20250626-0 (2025-06-27T11:35:39Z)

$ oc get mcp
NAME     CONFIG                                             UPDATED   UPDATING   DEGRADED   MACHINECOUNT   READYMACHINECOUNT   UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
master   rendered-master-f9398fc9986d7605b20e103c9bdc56a0   True      False      False      3              3                   3                     0                      4h10m
worker   rendered-worker-55c9b4e5fab5c47c17d76fc81f9170d7   False     True       False      2              1                   1                     0                      4h10m

When we remove the MachineOSConfig resource and the MCP is updated the MCP has to report the right number of nodes.