We did have some etcd latency last week, compacting and defragging the etcd instances stopped the etcd warnings, but this doesn't appear to have made a real difference