@Andrew This was the exact problem I was facing. The affected VMs still had the guest-tools.iso mounted in the virtual drive. Once I ejected the ISO, I was able to clear the errors and the backup jobs resumed normal operation.
For anyone else hitting this "stuck delta" or VDI_IN_USE state, here is the exact workflow I followed to resolve it:
Shut down the VM (Essential to release all storage locks).
Eject the ISO/CD from the VM console or CLI.
Delete all snapshots associated with the VM (including any old or "ghost" backup snapshots).
Delete the VM's backup folder from the remote SR/Backup target to ensure a clean metadata start.
Monitor the Coalesce process (via SMlog) and wait for it to finish flattening the VHD chain.
Power the VM back on.
Manually trigger a Backup Job.
Following these steps, the first run was (as expected) a Full backup, but every subsequent run has been a successful Delta with no "hanging" snapshots left behind.