I made a mistake when implementing T214493, by creating a long and weird data pipeline using EventLogging and ReportUpdater. The correct approach would be to simply emit StatsD metrics which can be directly consumed from Grafana to give all the same numbers. This leaves tech debt and if not communicated correctly, might accidentally slow down migration away from EventLogging, for example.
If we don't need the metrics at some point, let's remove the code and the data pipeline. If we decide to keep all or some of the metrics, let's rewrite using StatsD.