Add a dedicated serverStatus metric that records the duration of resharding critical section, rather than relying on transient active-phase timing. This should support monitoring in Grafana to identify prolonged write blocking. See the CS Dashboard doc for the monitoring gap.