kill-controller leaves instances with storage behind

WHen issuing kill-controller i expect it to issue termination commands with extreme prejudice and ordinance.

This doesn't appear to be happening on amazon when a unit has storage volumes attached. I can understand and respect juju not cleaning up the storage volumes as there is potential there for data loss, and we should do our best to preserve data volumes when we are aware of their existence

However, units which have storage attached dont seem to be getting cleaned up, as the volumes are never detached, which prevents kill-controller from actually terminating the unit, regardless of kill-controller, destroy-controller or otherwise.

Steps to reproduce:

juju deploy a service which supports storage (WIP using layer-docker)

juju add-storage docker/# disk-pool=ebs,,8G

chill while it attaches, formats, etc.

then issue a kill-controller to reset everything to zero, and log into the AWS control panel and view your instance still present, as well as any volumes you've provisioned.