Comment 4 for bug 1811726

I'd say this is medium/low severity since we do correlate compute_nodes / resource providers / ironic node resources via the same uuid, which comes from the ironic node uuid, so although we orphan any other ironic node resource providers in this case, if a compute service is managing those compute node records (ironic nodes) anywhere, it should re-use the same resource providers based on the same uuid. The possible downside is if you have the scheduler configured with a low value for limiting placement allocation candidates and placement is reporting resource providers that aren't actually being managed by any compute service - in that case you could get NoValidHost errors.