Rollback still fails to resolve retained image on RC13

[DeepZone]

Summary

Rollback still fails on Container Pilot 0.9.0-rc.13 when the retained image cannot be resolved through the stored repository/digest reference.

Environment

• Container Pilot: 0.9.0-rc.13
• Managed image: linuxserver/heimdall
• Docker host: self-hosted Linux Docker environment

Steps to reproduce

1. Update the Heimdall container through Container Pilot.
2. Confirm that the replacement container is running and a rollback point is shown.
3. Start the manual rollback.

Actual behavior

The rollback aborts with:

Docker GET /images/linuxserver%2Fheimdall%40sha256%3A3572daf882deca941fad00baecfa6755faedfb83ade4b6a6af761ad4e41b9244/json: 404 {"message":"No such image: linuxserver/heimdall@sha256:3572daf882deca941fad00baecfa6755faedfb83ade4b6a6af761ad4e41b9244"}

This still occurs after updating Container Pilot to RC13, which introduced local image-ID storage and digest-based fallback lookup.

Expected behavior

Container Pilot should resolve and restore the retained local image independently of its repository alias. If the image was actually removed, it should report that the local rollback image no longer exists without changing the running container.

Investigation notes

• Verify whether this RC12-created checkpoint contains an imageId.
• Inspect whether the retained/dangling image has an empty RepoDigests array.
• Consider persisting enough local image metadata to migrate legacy checkpoints that have neither a resolvable alias nor a discoverable RepoDigest.
• Add an integration case for a retained image whose original repository alias no longer resolves.

Acceptance criteria

• A rollback works when the previous image exists locally under a different alias or only by image ID.
• Legacy checkpoints receive a deterministic recovery path or a precise non-destructive error.
• Regression coverage reproduces the reported Docker 404.