feat(zulip): v3 resilience rewrite — circuit breaker, retry, watchdog, model fix

Root cause: Workers hung because default model was DeepSeek v4-pro (reasoning
model that produces empty content with normal token limits). pi RPC workers
waited forever for content that never arrived.

Changes:
- Circuit breaker (CLOSED→OPEN→HALF_OPEN) around Zulip API calls
- Retry with exponential backoff + jitter for transient errors
- Queue lifecycle management (idle_queue_timeout, auto re-register on BAD_EVENT_QUEUE_ID)
- External supervisor watchdog (restarts router after 3 health check failures)
- Busy worker timeout (SIGKILL after 5 min + error DM)
- PM2 hardening (max_restarts=100, max_memory_restart=500M)
- Crash handlers (uncaughtException + unhandledRejection → reconnect, not die)
- Fixed default model: deepseek-v4-pro → syslog-harness/syslog-auto
- Enhanced health endpoint with circuit breaker stats

Architecture: Research-backed from Zulip event system docs + Node.js resilience
patterns. Inline circuit breaker (no dependency). Separate watchdog process (Hermes pattern).

Verified: Circuit breaker trips on outage, recovers gracefully. End-to-end DM
processed in 4 seconds. Watchdog monitoring every 30s.

pi-zulip-extension/extension-src/watchdog.js

@@ -0,0 +1,84 @@
+/**
+ * Zulip Gateway Supervisor — external watchdog process.
+ *
+ * Monitors the router health endpoint every 30s. If 3 consecutive checks fail,
+ * restarts the abiba-zulip PM2 process gracefully.
+ *
+ * This is the pattern Hermes uses: an external supervisor that can recover
+ * the gateway even when the gateway process itself is hung (not just crashed).
+ *
+ * Deployed via PM2 as a separate process in ecosystem.config.cjs.
+ */
+
+const HEALTH_URL = "http://127.0.0.1:9200/health";
+const CHECK_INTERVAL_MS = 30_000;
+const MAX_FAILURES = 3;
+const RESTART_GRACE_MS = 15_000;
+
+let failures = 0;
+
+async function check() {
+  try {
+    const res = await fetch(HEALTH_URL, {
+      signal: AbortSignal.timeout(5000),
+    });
+
+    if (res.ok) {
+      const data = await res.json();
+
+      if (data.status === "ok" && data.zulip?.connected) {
+        if (failures > 0) {
+          console.log(`[watchdog] Router recovered after ${failures} failure(s)`);
+        }
+        failures = 0;
+        // Silent health log every 10 checks (~5 min) for monitoring
+        if (Math.random() < 0.1) {
+          console.log(`[watchdog] Router healthy (uptime: ${Math.round(process.uptime())}s)`);
+        }
+        return;
+      }
+
+      // Connected but degraded
+      console.warn(`[watchdog] Router degraded: status=${data.status}, connected=${data.zulip?.connected}`);
+      failures++;
+    } else {
+      console.warn(`[watchdog] Health check returned ${res.status}`);
+      failures++;
+    }
+  } catch (err) {
+    failures++;
+    console.warn(`[watchdog] Health check ${failures}/${MAX_FAILURES}: ${err.message}`);
+  }
+
+  if (failures >= MAX_FAILURES) {
+    console.error(`[watchdog] ${MAX_FAILURES} consecutive failures — restarting abiba-zulip`);
+
+    const { execSync } = await import("node:child_process");
+    try {
+      execSync("pm2 restart abiba-zulip", { timeout: 30000, encoding: "utf-8" });
+      console.log("[watchdog] Restart command sent successfully");
+    } catch (e) {
+      console.error(`[watchdog] Restart failed: ${e.message}`);
+
+      // Fallback: try resurrect if restart fails (process may be deleted)
+      try {
+        execSync("pm2 resurrect", { timeout: 30000 });
+        console.log("[watchdog] PM2 resurrected (fallback)");
+      } catch (e2) {
+        console.error(`[watchdog] Resurrect also failed: ${e2.message}`);
+      }
+    }
+
+    failures = 0;
+    // Wait for restart to fully initialize before checking again
+    await new Promise((r) => setTimeout(r, RESTART_GRACE_MS));
+  }
+}
+
+console.log("[watchdog] Zulip gateway supervisor started");
+console.log(`[watchdog] Monitoring ${HEALTH_URL} every ${CHECK_INTERVAL_MS / 1000}s`);
+console.log(`[watchdog] Max failures before restart: ${MAX_FAILURES}`);
+
+// Immediate first check, then periodic
+check();
+setInterval(check, CHECK_INTERVAL_MS);