Filebeat memory leak on Kubernetes?

We had other reasons to rebuild the K8s cluster. On restarting (and with Metricbeat now running as well) the Filebeat memory usage was flat for a couple of hours and then started increasing again - one pod restarted overnight.