Removing multiple running images is broken

[apostasie]

Description

When someimage is currently in use by a container, nerdctl rmi -f someimage will create a new unnamed image (name: :) so that we keep a ref on the layers (though this is currently buggy, see #4107).

But then, trying to remove multiple such images will fail, as image creation will bail out (image ":": already exists).

Does anyone have a suggestion on how to address this?
Is there a way to instruct containerd to create an image without a name? (eg: currently, passing en empty name will fail on creation)

Steps to reproduce the issue

nerdctl run -d debian sleep inf
nerdctl run -d alpine sleep inf
nerdctl rmi -f alpine
nerdctl rmi -f debian

Describe the results you received and expected

FATA[0000] 1 errors:
image ":": already exists

What version of nerdctl are you using?

main

Are you using a variant of nerdctl? (e.g., Rancher Desktop)

None

Host information

No response

[ogulcanaydogan]

Opened #5191. Same core idea as the earlier #5127 (name the dangling ref after its digest instead of the fixed ":" literal), verified end-to-end against a real containerd daemon: reproduced the original failure on unpatched main, then confirmed the fix on the branch.