The MCO guarantees that a certain class of MachineConfig updates (known in the MCO docs as "Rebootless Updates") does not cause the node to be rebooted. Currently, layered on-cluster builds cannot make this same guarantee because of how these files are applied to the underlying node (i.e., rpm-ostree rebase, then reboot into new image). Consequently, we'll need a way for this class of MachineConfigs to be updated on the node (likely in-situ) so that we can avoid rebooting the node for changes that only update that class of MachineConfigs.

As an implementation detail, it is possible that we can diff the current and desired MachineConfigs prior to beginning the layered update. And assuming nothing else has changed, we could then enter the usual MachineConfig update path to apply those updates as-is.

Done When:

• The MCD has rebootless update behavior parity within the on-cluster built image code path.