Startup Probe not honored in 4.8 since #RHBZ2095387

On Baremetal cluster with virtualizes control plane.
Customer deploys Springboot workloads with requests and limits (requests != limits).

When rebooting a node (config/upgrade) , since #RHBZ2095387, crio does not crash anymore but startup probes seems to be not honored.
Pods are starting all a the same time and startup timeout is not honored.

This results in pods killed before they started and be restarted and the high load of the some leads to workloads not starting of in a huge timeframe. 

4.10

Drain a node

1. Drain a node
2. Another node take the pods and issue occurs

500 pods star in ~ 2 days 

500 pods start in a reasonable time like before 4.8+

Also using TSP with 2 zones and anti affinity rules.