Looking for a solution for a while but couldn't find the best possible settings to sort the issue yet.

First of all, our application is lightweight and its resource consumption is very less.
We are trying to finetune the HPA settings of the deployment where we are facing the below issue.

1) During the new rollout, the deployment is creating multiple replicas (all are getting CPU % of 80 above) using the hpa for a short time and scaling down after some time. - What is the factor which makes the hpa to autoscale only during the new rollout - What is the reason for making high CPU consumption in the pods just during the rollout (creating time) and which in turn increasing the replicas. But normal usage of the pods are very very less

So as a solution looking for the best hpa properties to enable, - so that I can control the unwanted auto pod scaling during the new rollouts considering the below properties.

               - policy window (period)
               - stabilization window
               - CPU initialization period
               - initial readiness

How I can finetune the HPA with the above parameters to resolve the issue.