If OpenShift Pipelines is used in combination with Cluster Autoscaler, it will probably happen that some pipeline tasks will be interrupted because their pods have become OutOfMemory or OutOfCpu.

It would be optimal to force all the pipeline task pods to stay in Pending status until a new node is deployed by the Autoscaler instead of being scheduled, becoming OutOfCpu (for example) and consequently interrupted.

There is already SRVKP-2434 where it is being requested that the QoS of all the task pods can be set to Guaranteed. Would that feature be enough to achieve this request or would it be needed to develop anything else? In the latter case, the request here would be that the additional necessary implementations for this are carried out.