When using the LifecycleAndUtilization profile, the descheduler identifies overutilized nodes and evicts pods without checking whether a suitable destination node exists. As a result, pods are frequently moved between nodes that are all equally overutilized, leading to unnecessary eviction churn and workload instability.
Implementing target-node filtering (the nodeFit feature) in the Kube Descheduler Operator would allow users to configure the descheduler to check for valid destination nodes before evicting pods. This ensures more efficient and stable descheduling decisions.