Whereabouts currently can leak IP addresses when a pod is terminated forcefully (e.g. "oc delete pod foo --force --grace-period=0") or when a node is shutdown forcefully. The Whereabouts CNI does

A couple possible solutions:

• A k8s cron job that periodically scans the IP address allocations and checks for the presence of a pod
• A k8s controller that watches the IP address allocations

My preference would be a cron job and later expand to a full controller.