keep_os_disk appears to conflict with packer deleting the resource group?

Over in https://github.com/hashicorp/packer/issues/6752, there was a request to allow copying the image to another subscription.
PR https://github.com/hashicorp/packer/pull/10045 added a `keep_os_disk` flag as a workaround to allow the use of `az image copy` as a post build step.

I've added `keep_os_disk = true`, but it appears that the disk is deleted when the (packer created) resource group is deleted.

How is this meant to work? Does it only work if you provide a pre-created resource group?

ps - I would have continued the discussion on the issue or the PR, but they've unhelpfully been locked.

Start by reading the keep_os_disk behavior described here alongside Packer issue 6752 and pull request 10045, then trace how the packer-created resource group is handled. Determine whether the disk is expected to survive resource-group deletion or requires a pre-created resource group, and document or reproduce the result with az image copy.

Written by the indexing model from the issue text.