fix: ensure scheduler starts only once during Flask requests

tests/test_scheduler.py

@@ -153,3 +153,23 @@ class TestScheduler:
         assert search_data.get(("losangeles", "python")) == 4
         assert search_data.get(("losangeles", "java")) == 1
         assert result.get("discovered") == 10  # Total unique jobs
+
+
+def test_app_scheduler_starts_once(monkeypatch):
+    """Ensure the Flask before_request hook starts scheduler only once."""
+    import web.app as app_module
+
+    monkeypatch.setenv("SCRAPE_SCHEDULER_ENABLED", "true")
+    monkeypatch.delenv("SERVER_SOFTWARE", raising=False)
+    monkeypatch.delenv("FLASK_RUN_FROM_CLI", raising=False)
+    monkeypatch.delenv("WERKZEUG_RUN_MAIN", raising=False)
+
+    app_module._scheduler_started = False
+
+    with patch("web.app.start_scheduler_in_background") as mock_start:
+        app_module.app.config.update(TESTING=True, WTF_CSRF_ENABLED=False)
+        with app_module.app.test_client() as client:
+            client.get("/health")
+            client.get("/health")
+
+        assert mock_start.call_count == 1