diff --git a/tubesync/sync/tasks.py b/tubesync/sync/tasks.py
index d004362a..7e7a0f0f 100644
--- a/tubesync/sync/tasks.py
+++ b/tubesync/sync/tasks.py
@@ -181,7 +181,6 @@ def cleanup_removed_media(source, videos):
 
 
 @background(schedule=300, remove_existing_tasks=True)
-@atomic(durable=True)
 def index_source_task(source_id):
     '''
         Indexes media available from a Source object.
@@ -206,51 +205,54 @@ def index_source_task(source_id):
     source.save()
     log.info(f'Found {len(videos)} media items for source: {source}')
     fields = lambda f, m: m.get_metadata_field(f)
-    for video in videos:
-        # Create or update each video as a Media object
-        key = video.get(source.key_field, None)
-        if not key:
-            # Video has no unique key (ID), it can't be indexed
-            continue
-        try:
-            media = Media.objects.get(key=key, source=source)
-        except Media.DoesNotExist:
-            media = Media(key=key)
-        media.source = source
-        media.duration = float(video.get(fields('duration', media), None) or 0) or None
-        media.title = str(video.get(fields('title', media), ''))[:200]
-        timestamp = video.get(fields('timestamp', media), None)
-        published_dt = media.metadata_published(timestamp)
-        if published_dt is not None:
-            media.published = published_dt
-        try:
-            with atomic():
-                media.save()
-            log.debug(f'Indexed media: {source} / {media}')
-            # log the new media instances
-            new_media_instance = (
-                media.created and
-                source.last_crawl and
-                media.created >= source.last_crawl
-            )
-            if new_media_instance:
-                log.info(f'Indexed new media: {source} / {media}')
-                log.info(f'Scheduling task to download metadata for: {media.url}')
-                verbose_name = _('Downloading metadata for "{}"')
-                download_media_metadata(
-                    str(media.pk),
-                    priority=9,
-                    verbose_name=verbose_name.format(media.pk),
+    with atomic(durable=True):
+        for video in videos:
+            # Create or update each video as a Media object
+            key = video.get(source.key_field, None)
+            if not key:
+                # Video has no unique key (ID), it can't be indexed
+                continue
+            try:
+                media = Media.objects.get(key=key, source=source)
+            except Media.DoesNotExist:
+                media = Media(key=key)
+            media.source = source
+            media.duration = float(video.get(fields('duration', media), None) or 0) or None
+            media.title = str(video.get(fields('title', media), ''))[:200]
+            timestamp = video.get(fields('timestamp', media), None)
+            published_dt = media.metadata_published(timestamp)
+            if published_dt is not None:
+                media.published = published_dt
+            try:
+                with atomic():
+                    media.save()
+            except IntegrityError as e:
+                log.error(f'Index media failed: {source} / {media} with "{e}"')
+            else:
+                log.debug(f'Indexed media: {source} / {media}')
+                # log the new media instances
+                new_media_instance = (
+                    media.created and
+                    source.last_crawl and
+                    media.created >= source.last_crawl
                 )
-        except IntegrityError as e:
-            log.error(f'Index media failed: {source} / {media} with "{e}"')
+                if new_media_instance:
+                    log.info(f'Indexed new media: {source} / {media}')
+                    log.info(f'Scheduling task to download metadata for: {media.url}')
+                    verbose_name = _('Downloading metadata for "{}"')
+                    download_media_metadata(
+                        str(media.pk),
+                        priority=20,
+                        verbose_name=verbose_name.format(media.pk),
+                    )
     # Tack on a cleanup of old completed tasks
     cleanup_completed_tasks()
-    # Tack on a cleanup of old media
-    cleanup_old_media()
-    if source.delete_removed_media:
-        log.info(f'Cleaning up media no longer in source: {source}')
-        cleanup_removed_media(source, videos)
+    with atomic(durable=True):
+        # Tack on a cleanup of old media
+        cleanup_old_media()
+        if source.delete_removed_media:
+            log.info(f'Cleaning up media no longer in source: {source}')
+            cleanup_removed_media(source, videos)
 
 
 @background(schedule=0)