fix(supervisor): sweep stale activeCalls before returning to Live Monitor

Bug 560: Live Call Monitor showed ghost calls with runaway timers when
the agent wasn't on a call. Cause — activeCalls Map only added on
'Answered' and deleted on 'Disconnect'; a missed Disconnect (sidecar
restart, Ozonetel subscription hiccup, network blip) left the entry
lingering forever.

getActiveCalls() now sweeps stale entries before returning:
 - drop if startTime is older than 30 minutes
 - drop if the mapped agent is currently ready / offline / paused
   (agent can't be on a call in any of those states)

Each sweep logs the reason so we can track how often this fires.

src/supervisor/supervisor.service.ts

@@ -257,7 +257,52 @@ export class SupervisorService implements OnModuleInit {
         this.agentStateSubject.next({ agentId, state: 'force-logout' as any, timestamp: new Date().toISOString() });
     }
 
+    // Max plausible call length before the entry is treated as orphaned.
+    // Real Ozonetel calls cap out far short of this — 30 minutes is a safe
+    // ceiling for a hospital call-center context. If a genuinely longer
+    // call existed, losing it from Live Monitor is preferable to the ghost
+    // state (supervisors lose trust in the dashboard otherwise).
+    private static readonly MAX_ACTIVE_CALL_AGE_MS = 30 * 60 * 1000;
+
+    // Agent states that are incompatible with having an active call. If the
+    // mapped agent is currently in one of these, the activeCalls entry is
+    // definitely stale (e.g. Disconnect webhook was dropped).
+    private static readonly NON_CALL_AGENT_STATES = new Set(['ready', 'offline', 'paused']);
+
     getActiveCalls(): ActiveCall[] {
+        // Sweep stale entries before returning. The activeCalls Map is a
+        // best-effort in-memory projection of Ozonetel call events — if
+        // Ozonetel drops a Disconnect (network blip, subscription hiccup,
+        // sidecar restart mid-call), the entry lingers forever and the
+        // Live Call Monitor shows a ghost call with a runaway timer.
+        //
+        // Two signals identify staleness:
+        //   1. The associated agent is not in a busy state (ready, offline,
+        //      paused — they can't be on a call).
+        //   2. startTime is older than MAX_ACTIVE_CALL_AGE_MS (hard ceiling
+        //      regardless of agent-state signal).
+        const now = Date.now();
+        const toDelete: string[] = [];
+
+        for (const [ucid, call] of this.activeCalls.entries()) {
+            const ageMs = now - new Date(call.startTime).getTime();
+            if (isNaN(ageMs)) continue;
+
+            if (ageMs > SupervisorService.MAX_ACTIVE_CALL_AGE_MS) {
+                toDelete.push(ucid);
+                this.logger.warn(`[ACTIVE-CALLS] Sweep: dropping ${ucid} (age ${Math.round(ageMs / 60000)}m, exceeds ${SupervisorService.MAX_ACTIVE_CALL_AGE_MS / 60000}m cap)`);
+                continue;
+            }
+
+            const agentState = this.agentStates.get(call.agentId)?.state;
+            if (agentState && SupervisorService.NON_CALL_AGENT_STATES.has(agentState)) {
+                toDelete.push(ucid);
+                this.logger.warn(`[ACTIVE-CALLS] Sweep: dropping ${ucid} — agent ${call.agentId} is ${agentState}`);
+            }
+        }
+
+        for (const ucid of toDelete) this.activeCalls.delete(ucid);
+
         return Array.from(this.activeCalls.values());
     }