High replag is a common symptom, but we think that different shapes of workloads induce replag differently. To aid us in incidents and in future development, we should identify metrics that we can capture to guide both remediations and improvements.