diff --git a/backend/app/services/scanner.py b/backend/app/services/scanner.py
index f55ac25..3aa9b07 100644
--- a/backend/app/services/scanner.py
+++ b/backend/app/services/scanner.py
@@ -8,7 +8,7 @@ from sqlalchemy import select
 
 from app.database import AsyncSessionLocal
 from app.models import SourceRoot
-from app.tasks.scan import scan_all_source_roots, watch_folders
+from app.tasks.scan import scan_all_source_roots
 from app.config import settings
 
 logger = logging.getLogger(__name__)
@@ -49,15 +49,17 @@ async def bootstrap_default_source_root() -> None:
 
 
 async def start_initial_scan():
-    """Start the initial library scan"""
+    """Start the initial library scan.
+
+    NOTE: the folder watcher (watch_folders task) is intentionally NOT
+    dispatched here. It's an infinite loop celery task and every backend
+    restart was queuing a new instance, eventually pinning every worker
+    and starving scan_folder dispatches. Re-enabling it needs a Redis
+    lock or a dedicated long-running container — until then the user
+    triggers scans manually via "Scan all folders".
+    """
     try:
-        # Queue scan of all source roots
         scan_all_source_roots.delay()
-
-        # Start folder watcher if configured
-        if settings.scanner.watch:
-            watch_folders.delay()
-
         logger.info("Initial scan queued successfully")
     except Exception as e:
         logger.error(f"Failed to start initial scan: {e}")