In order to safely roll out the new method of calculating the cpu system reserved values [1] , we would have to introduce versioning in auto node sizing. This way even if the new method ends up reserving more CPU, existing customers won't see any dip in the amount of CPU available for their workloads. 

[1] https://issues.redhat.com/browse/OCPNODE-2211

https://issues.redhat.com/browse/OCPNODE-2226 introduced the versioning mechanism in 4.17. This issue is to create a feature backport for 4.16