Metrics and dashboards for the new platform

The measurement that predicted the storage stalls behind this whole migration is still not collected. Wiring it in, with an alert on the specific disk-latency figure, is what stops the next occurrence being discovered weeks later.