Comment 3 for bug 1600031

Actually I'm thinking it's not https://review.openstack.org/#/c/334094/ because that's just changing up a bit how we create the instance and BDMs in the DB from the API, at that point we haven't even hit the scheduler or built the instance on the compute, which means for the periodic task in the compute to fail we have to have guests with those instance uuids, which wouldn't be at this point, so this is more likely a race with delete, so now I'm thinking it's related to https://review.openstack.org/#/c/335697/ .