diff --git a/database/manager.py b/database/manager.py
index 5706800cc..a902c7bcc 100644
--- a/database/manager.py
+++ b/database/manager.py
@@ -1243,7 +1243,7 @@ def _schedule_background_reconnect(
         kwargs: Dict[str, Any],
         mongo_url: Optional[str],
         database_name: Optional[str],
-        delay: float = 30.0,
+        delay: float = 5.0,
         max_bg_attempts: int = 10,
         _attempt: int = 1,
     ) -> None:
@@ -1293,13 +1293,24 @@ def _try_reconnect():
                     client.close()
                 except Exception:
                     pass
-                if _attempt < max_bg_attempts:
-                    next_delay = min(delay * 1.5, 300.0)
-                    self._schedule_background_reconnect(
-                        kwargs, mongo_url, database_name,
-                        delay=next_delay, max_bg_attempts=max_bg_attempts,
-                        _attempt=_attempt + 1,
+                # Never give up: main.py now does a passive wait instead of SystemExit,
+                # so if we stop scheduling here the app would spin forever without any
+                # active reconnect attempt. Exponential backoff is capped at 300s, so
+                # load stays bounded. We emit a one-time escalation event when the
+                # historical max is crossed, for observability.
+                if _attempt == max_bg_attempts:
+                    emit_event(
+                        "db_background_reconnect_escalating",
+                        severity="error",
+                        attempt=_attempt,
+                        note="continuing with capped backoff",
                     )
+                next_delay = min(delay * 1.5, 300.0)
+                self._schedule_background_reconnect(
+                    kwargs, mongo_url, database_name,
+                    delay=next_delay, max_bg_attempts=max_bg_attempts,
+                    _attempt=_attempt + 1,
+                )
 
         timer = threading.Timer(delay, _try_reconnect)
         timer.daemon = True
diff --git a/docs/environment-variables.rst b/docs/environment-variables.rst
index 93dd71c1e..b8f70ca8c 100644
--- a/docs/environment-variables.rst
+++ b/docs/environment-variables.rst
@@ -406,6 +406,18 @@
      - ``1000``
      - ``1500``
      - Bot/WebApp
+   * - ``DB_RECONNECT_WAIT_BEFORE_POLL``
+     - זמן המתנה ראשוני (שניות) בעלייה של הבוט עד שה-DB חוזר, לפני שעוברים ל-poll פסיבי. הבוט לא יוצא (SystemExit) אחרי החלון — הוא ממשיך להמתין.
+     - לא
+     - ``120``
+     - ``120``
+     - Bot
+   * - ``DB_RECONNECT_POLL_INTERVAL``
+     - מרווח (שניות) בין בדיקות חיבור ב-poll פסיבי לאחר שחלון ההמתנה הראשוני פג. ה-reconnect ברקע ממשיך לנסות עם backoff מעריכי (מקסימום 300 שניות).
+     - לא
+     - ``30``
+     - ``30``
+     - Bot
    * - ``DB_HEALTH_POOL_REFRESH_SEC``
      - תדירות רענון מומלצת (שניות) לסטטוס ה-pool בדשבורד. (משתנה תיעודי/קונפיגורציה כללית)
      - לא
diff --git a/main.py b/main.py
index 7df7eb286..1216a454d 100644
--- a/main.py
+++ b/main.py
@@ -4917,9 +4917,20 @@ def main() -> None:
                 time.sleep(5)
                 _db_waited += 5
             if not getattr(db, 'is_connected', True):
-                logger.critical("DB still unreachable after %ds; exiting to prevent unsafe operation.", _db_wait_max)
-                raise SystemExit(1)
-            logger.info("DB reconnected after %ds wait; proceeding with lock acquisition.", _db_waited)
+                # Passive wait instead of SystemExit: keep the process alive and let the
+                # background reconnect keep trying. Avoids CrashLoop + Sentry storms when
+                # the DB has a transient outage.
+                _poll_interval = int(os.getenv("DB_RECONNECT_POLL_INTERVAL", "30"))
+                logger.warning(
+                    "DB still unreachable after %ds; entering passive wait (polling every %ds).",
+                    _db_wait_max, _poll_interval,
+                )
+                while not getattr(db, 'is_connected', True):
+                    time.sleep(_poll_interval)
+                    _db_waited += _poll_interval
+                logger.info("DB reconnected after %ds total wait; proceeding with lock acquisition.", _db_waited)
+            else:
+                logger.info("DB reconnected after %ds wait; proceeding with lock acquisition.", _db_waited)
 
         # MongoDB connection and lock management
         if not manage_mongo_lock():
diff --git a/services/config_inspector_service.py b/services/config_inspector_service.py
index b066e5dcc..354b7a6a6 100644
--- a/services/config_inspector_service.py
+++ b/services/config_inspector_service.py
@@ -189,6 +189,18 @@ class ConfigService:
             category="database",
             sensitive=True,
         ),
+        "DB_RECONNECT_WAIT_BEFORE_POLL": ConfigDefinition(
+            key="DB_RECONNECT_WAIT_BEFORE_POLL",
+            default="120",
+            description="זמן המתנה ראשוני (שניות) להתחברות מחדש ל-DB בעלייה לפני מעבר ל-poll פסיבי",
+            category="database",
+        ),
+        "DB_RECONNECT_POLL_INTERVAL": ConfigDefinition(
+            key="DB_RECONNECT_POLL_INTERVAL",
+            default="30",
+            description="מרווח (שניות) בין בדיקות חיבור ב-poll פסיבי לאחר שחלון ההמתנה הראשוני פג",
+            category="database",
+        ),
         "DB_HEALTH_SLOW_THRESHOLD_MS": ConfigDefinition(
             key="DB_HEALTH_SLOW_THRESHOLD_MS",
             default="1000",