We have a case where a user:
- Create a 3+3 cluster
- Creates a config patch like:
machine:
nodeLabels:
invalid label: randomvalue
- Update it to be valid:
machine:
nodeLabels:
valid-label: randomvalue
Try it a few times, use multiple patches and so on - it should eventually get into the state where there is no invalid label in any of the patches, but it still is present in the machine config, and the config error visible on the UI.
(Might be related to the resource cache).