fix(service): skip retained history for unavailable monitors at startup

Ignore history rows without initialized daily or monthly state so deleted and quarantined services cannot panic dashboard startup. Cover startup and statistics preservation with regression tests.

Thanks to Zachariah Mithani (@zachariah-mithani, AploSec) for privately reporting the retained service history startup panic.

Co-authored-by: naiba/CloudCode <hi+cloudcode@nai.ba>

service/singleton/servicesentinel.go

@@ -431,10 +431,18 @@ func (ss *ServiceSentinel) loadTodayStats(today time.Time) {
 		totalDelay := make(map[uint64]float64)
 		totalDelayCount := make(map[uint64]int)
 		for _, mh := range mhs {
-			ss.serviceStatusToday[mh.ServiceID].Up += mh.Up
-			ss.monthlyStatus[mh.ServiceID].TotalUp += mh.Up
-			ss.serviceStatusToday[mh.ServiceID].Down += mh.Down
-			ss.monthlyStatus[mh.ServiceID].TotalDown += mh.Down
+			todayStats := ss.serviceStatusToday[mh.ServiceID]
+			monthlyStats := ss.monthlyStatus[mh.ServiceID]
+			// History can outlive a deleted service or belong to a quarantined
+			// service. Neither has startup status entries; skip its rows before
+			// accumulating counts or delays so retained history cannot panic startup.
+			if todayStats == nil || monthlyStats == nil {
+				continue
+			}
+			todayStats.Up += mh.Up
+			monthlyStats.TotalUp += mh.Up
+			todayStats.Down += mh.Down
+			monthlyStats.TotalDown += mh.Down
 			totalDelay[mh.ServiceID] += mh.AvgDelay
 			totalDelayCount[mh.ServiceID]++
 		}