Glenn Matthews

Thanks for the report. I'm honestly a bit confused - are you saying that in this case, when worker 2 runs the job, it's somehow running the `2f071d5` version of...

Thing I'm still struggling with is that Nautobot only cares about what code is currently checked out, it shouldn't *matter* whether that's actually seen by Git as a branch or...

Based on the outcomes of #5164, let's regroup and see if we can instead figure out a way to get our Celery worker (possibly by overriding [`on_task_received`](https://github.com/celery/celery/blob/main/celery/worker/consumer/consumer.py#L641)?) to forcibly re-import...

Reproducible under `develop` by the following procedure: 1. Edit `development/docker-compose.yml` to create three replicas of the `celery_worker` service. 1. Create a GitRepository pointed to https://github.com/nautobot/demo-git-datasource.git, branch `nautobot-jobs` (or `main`), and...

May be related to https://github.com/celery/celery/pull/8491#issuecomment-2119191201

Might be fixed in Celery 5.4.1 when released, see https://github.com/celery/celery/pull/9038

Looks like the fix will be in Celery 5.5 at this point.

Probably missing logic in `DynamicGroupEditView.post()` since it overrides the base `ObjectEditView.post()` implementation.

ObjectChange records are another model that make extensive use of GFKs and will be affected by this issue.

Initial prototype at https://github.com/nautobot/nautobot/tree/prototype/3666-serialize-deserialize-genericforeignkey; further progress and validation is blocked by the lack of fully sane natural keys on all core models (see #3752 and child issues)