Comment 3 for bug 1767363

_disconnect_volume doesn't delete host volume attachments so the worst thing that can happen with this race is that the underlying volume connection isn't cleaned up on the host when multiple requests race each other. I'll throw a volume_id specific lock around the method to avoid this.

FWIW c#1 and c#2 isn't related, that shows a failure to detach resulting in n-cpu rolling the detach back in c-api.