Q. How do we get of a pipeline-cancel situation? It keeps going round-and-round but does not come back.

Answer:

a)  Please delete the underlying objects in k8s/deployment envrionment. For example, if we were trying to create a "deployment", go and delete it. That should fix the issue. That object, if it is crashloop, will not go away easily. A simple approach is to have pipelines that delete objects created by other pipelines...so you can simple run a pipeline to clean-up a deployment left in a messy state.

b) orca pod could have run out of resources. Try restarting it or increasing its CPU/Mem.