clean stale dispatcher connections closer to post-fork

see: https://github.com/ansible/awx/issues/9559

awx/conf/settings.py

@@ -352,7 +352,6 @@ class SettingsWrapper(UserSettingsHolder):
         self.cache.set_many(settings_to_cache, timeout=SETTING_CACHE_TIMEOUT)
 
     def _get_local(self, name, validate=True):
-        self.__clean_on_fork__()
         self._preload_cache()
         cache_key = Setting.get_cache_key(name)
         try:

awx/main/dispatch/worker/task.py

@@ -6,6 +6,7 @@ import traceback
 
 from kubernetes.config import kube_config
 
+from django.conf import settings
 from django_guid.middleware import GuidMiddleware
 
 from awx.main.tasks import dispatch_startup, inform_cluster_of_shutdown
@@ -85,6 +86,7 @@ class TaskWorker(BaseWorker):
             'task': u'awx.main.tasks.RunProjectUpdate'
         }
         '''
+        settings.__clean_on_fork__()
         result = None
         try:
             result = self.run_callable(body)