Doesn't validate the existing nodes have enough resources

[vaibhavkhurana2018]

While calculating for a better node it doesn't account for the allocatable resources on the node. So if the resources are not available on the existing nodes it will go on an infinite loop as the pods won't be fitting in the nodes and will be rescheduled by the k8s scheduler back on the current nodes.

[vaibhavkhurana2018]

From my understanding if we can filter the preferred nodes with PodFitsResources(https://pkg.go.dev/knative.dev/eventing/pkg/scheduler/plugins/core/podfitsresources#pkg-types) filter and then take the decision to evict the pods, should be a good start on solving the issue.