diff --git a/CHANGELOG.md b/CHANGELOG.md
index 24203d49..0daba010 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -39,6 +39,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Version update notification persisting after upgrading to the notified version (e.g. "v0.20.2 available" shown while already running v0.20.2). Root cause: `check_for_version_update.delay()` was called from `AppConfig.ready()`, which fires inside Celery prefork pool subprocesses before the broker connection is established, causing the dispatch to fail silently with no log output. Fixed by moving the startup dispatch to the `worker_ready` signal in `celery.py` (consistent with the existing `recover_recordings_on_startup` pattern), and deleting the stale `version-{current_version}` notification at the top of the production check path so it is cleared even when GitHub is unreachable. A WebSocket update is sent immediately on deletion so the frontend badge clears without waiting for the API response.
 - VOD orphan cleanup crashing with a `ForeignKeyViolation` (`IntegrityError`) when a concurrent refresh task created a new `M3UMovieRelation` or `M3USeriesRelation` for a movie/series between the orphan-detection query and the `DELETE` SQL. Both `orphaned_movies.delete()` and `orphaned_series.delete()` are now wrapped in `try/except IntegrityError`; affected records are skipped with a warning and will be cleaned up on the next scheduled run.
 - XC stream refresh crashing with a `null value in column "name"` database error when a provider returns streams with a null or empty name. Affected streams are now assigned a generated fallback name in the format `<account name> - <stream_id>` so the refresh completes successfully and the stream remains accessible. A warning is logged for each affected stream.
 - 504 Gateway Timeout when saving M3U group settings on slower hardware (e.g. Synology NAS). Replaced per-row `update_or_create()` loops with `bulk_create(update_conflicts=True)` wrapped in `transaction.atomic()` for both `ChannelGroupM3UAccount` and `M3UVODCategoryRelation`, reducing hundreds of individual DB round-trips to a single query per model. (Fixes #745) — Thanks [@nickgerrer](https://github.com/nickgerrer)
diff --git a/core/apps.py b/core/apps.py
index f2780bd1..ee182e89 100644
--- a/core/apps.py
+++ b/core/apps.py
@@ -31,7 +31,6 @@ class CoreConfig(AppConfig):
             return
 
         self._sync_developer_notifications()
-        self._check_version_update()
 
     def _sync_developer_notifications(self):
         """Sync developer notifications from JSON file to database."""
@@ -47,14 +46,3 @@ class CoreConfig(AppConfig):
         except Exception as e:
             logger.warning(f"Failed to sync developer notifications on startup: {e}")
 
-    def _check_version_update(self):
-        """Check for version updates on startup."""
-        import logging
-
-        logger = logging.getLogger(__name__)
-
-        try:
-            from core.tasks import check_for_version_update
-            check_for_version_update.delay()
-        except Exception as e:
-            logger.warning(f"Failed to check for version updates on startup: {e}")
diff --git a/core/tasks.py b/core/tasks.py
index 44ddc68a..a18c2f74 100644
--- a/core/tasks.py
+++ b/core/tasks.py
@@ -789,6 +789,7 @@ def check_for_version_update():
     try:
         is_dev_build = __timestamp__ is not None
         DISPATCHARR_HEADERS = {'User-Agent': f'Dispatcharr/{__version__}'}
+
         if is_dev_build:
             # Check Docker Hub for newer dev builds
             docker_hub_url = "https://hub.docker.com/v2/repositories/dispatcharr/dispatcharr/tags/dev"
@@ -878,7 +879,21 @@ def check_for_version_update():
                         }
                     )
         else:
-            # Production build - check GitHub for stable releases
+            # Production build - check GitHub for stable releases.
+            # Delete any stale notification for the currently running version upfront;
+            # a "vX is available" notification is meaningless once the user is already on vX.
+            # Notify the frontend immediately so the badge clears without waiting for the API call.
+            deleted_count = SystemNotification.objects.filter(
+                notification_key=f"version-{__version__}",
+                notification_type='version_update',
+            ).delete()[0]
+            if deleted_count > 0:
+                send_websocket_update(
+                    'updates',
+                    'update',
+                    {'success': True, 'type': 'notifications_cleared', 'count': deleted_count}
+                )
+
             github_api_url = "https://api.github.com/repos/Dispatcharr/Dispatcharr/releases/latest"
             headers = {"Accept": "application/vnd.github.v3+json", **DISPATCHARR_HEADERS}
             response = requests.get(
diff --git a/dispatcharr/celery.py b/dispatcharr/celery.py
index 436ffc35..8ccb3c33 100644
--- a/dispatcharr/celery.py
+++ b/dispatcharr/celery.py
@@ -54,7 +54,7 @@ app.conf.update(
 def cleanup_task_memory(**kwargs):
     """Clean up memory and database connections after each task completes"""
     from django.db import connection
-    
+
     # Get task name from kwargs
     task_name = kwargs.get('task').name if kwargs.get('task') else ''
 
@@ -153,6 +153,9 @@ def setup_celery_logging(**kwargs):
 
 @worker_ready.connect
 def on_worker_ready(**kwargs):
-    """Resume or finalize interrupted DVR recordings after a worker restart."""
+    """Tasks to run once the worker is fully connected and ready."""
     from apps.channels.tasks import recover_recordings_on_startup
     recover_recordings_on_startup.delay()
+
+    from core.tasks import check_for_version_update
+    check_for_version_update.delay()