To prevent the rolling of all brokers pods, or the unanticipated disruption of other pods, such as ingress replicas, we need to account for how scheduling changes are applied in a migration scenario.

For controlling broker rolling, it has been proposed to use an artificial strimzi version to control roll the instances. This will be controlled by a dynamic-scaling-scheduling flag in the strimzi configmap in the fleetshard_operands.yaml.

For ingress replicas we need them to roll onto the default machine pool before adding the NoExecute taint. For this we'll need to also use the NoSchedule version of the taint, and update the existing tolerations to include that.