Updating the ApplicationGroup after a successful reconcile sometimes causes the updated workflow to never run

1. Deploy an applicationgroup and monitor the status till it switches to `Succeeded`.
2. Edit the applicationgroup and reapply it

The workflow status shows that it is running, but no pods are seen being spun up for the workflow.

```console
> kubectl get workflow
NAME STATUS AGE
dev Running 4m42s
```

```log
orkestra-workflow-controller-5c66ffdf88-6rf59 controller time="2021-03-17T03:05:22.396Z" level=info msg="Pod node dev-584293088 initialized Pending" namespace=orkestra workflow=dev
orkestra-workflow-controller-5c66ffdf88-6rf59 controller time="2021-03-17T03:05:22.413Z" level=info msg="Failed pod dev.redis-dev.redis-dev (dev-584293088) creation: already exists" namespace=orkestra workflow=dev
orkestra-workflow-controller-5c66ffdf88-6rf59 controller time="2021-03-17T03:05:22.420Z" level=info msg="Workflow update successful" namespace=orkestra phase=Running resourceVersion=4456 workflow=dev
```

This is possibly an argo issue but we need a way to remediate this error.