fix: harden high-fanout delivery under load

server/metrics/metrics.go

@@ -29,6 +29,10 @@ var (
 		Help:    "Database operation duration by driver, operation, and status.",
 		Buckets: []float64{0.001, 0.003, 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5},
 	}, []string{"driver", "operation", "status"})
+	DBInvalidConnectionRetryTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
+		Name: "im_db_invalid_connection_retry_total",
+		Help: "Retries after an invalid database connection by operation and outcome.",
+	}, []string{"operation", "outcome"})
 
 	RealtimeFanoutEventsTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
 		Name: "realtime_fanout_events_total",
@@ -228,7 +232,7 @@ func init() {
 		IMWSSendQueueCurrent, IMWSSendQueueCapacity,
 		IMWSSendQueueEnqueueTotal, IMWSSendQueueDepth, IMWSSendQueueWaitDuration,
 		IMClusterLivePeers, IMClusterExpectedPeers,
-		HTTPRequestsTotal, HTTPRequestDuration, DBQueryDuration,
+		HTTPRequestsTotal, HTTPRequestDuration, DBQueryDuration, DBInvalidConnectionRetryTotal,
 		RealtimeFanoutEventsTotal, RealtimeFanoutDeliveryDuration, ClusterP2MQueueLength,
 		ClusterP2MQueueCapacity, ClusterP2MEnqueueTotal, ClusterRPCInflight, ClusterRPCDuration,
 		ClusterRouteTotal, ClusterProxyResponseTotal, InternalQueueEnqueueTotal, InternalQueueDepth,
@@ -345,6 +349,10 @@ func IncWSHeartbeat(event string) {
 	IMWSHeartbeatTotal.WithLabelValues(metricPodName(), normalizeLabel(event)).Inc()
 }
 
+func IncDBInvalidConnectionRetry(operation, outcome string) {
+	DBInvalidConnectionRetryTotal.WithLabelValues(normalizeLabel(operation), normalizeLabel(outcome)).Inc()
+}
+
 func SetWSHeartbeatConfig(pingPeriod, pongWait time.Duration) {
 	pod := metricPodName()
 	IMWSHeartbeatSeconds.WithLabelValues(pod, "ping_period").Set(pingPeriod.Seconds())