bootc status/upgrade fails after upgrade with ostree-finalize-staged.service exceeding timeout, rendering bootc unusable

I have been enjoying using Fedora bootc for quite a while now in my homelab :)

Occasionally after updating one of my systems (image based on fedora-bootc:42) I get this error message from bootc upgrade (and bootc status):

```shell
$ sudo bootc upgrade
error: Upgrading: Computing status: Rollback deployment: Reading deployment metadata: Missing base image ref ostree/container/blob/sha256_3A_0b11f38bc7a328029152915418e3d218669bacef4aafafd79e6501d7c4622e9b
```

It seems the missing ref is always from the rollback image and the error occurs if the `ostree-finalize-staged.service` times out after a previous upgrade. The default timeout is set to 5min:

```ini
# /usr/lib/systemd/system/ostree-finalize-staged.service
[Service]
# This is a quite long timeout intentionally; the failure mode
# here is that people don't get an upgrade. We need to handle
# cases with slow rotational media, etc.
TimeoutStopSec=5m
```

Evidently this is not enough for my arm SBC (librecomputer lafrite) with slow USB storage. After overwriting it with `TimeoutStopSec=10m` I have not encountered this issue for 3 months.

But unlike what the comment in the service file suggests is does not lead to the upgrade failing, but in **rendering the system unable to upgade, rollback or applying another image**.

Unfortunately I fond no documentation describing how to recover from this error (only #219, seems related?). Instead I managed to recover my system using the undocumented internals commands, but is there a better/intended way?

```shell
$ sudo bootc upgrade
error: Upgrading: Computing status: Rollback deployment: Reading deployment metadata: Missing base image ref ostree/container/blob/sha256_3A_0b11f38bc7a328029152915418e3d218669bacef4aafafd79e6501d7c4622e9b

$ sudo mount -o remount,rw /sysroot
# pull the missing (?) rollback image
$ sudo bootc internals ostree-ext container image pull /sysroot/ostree/repo ostree-unverified-registry:registry.local/bootc/lafrite-bootc@sha256:a3318fb5cd371288b77a64f3177e9b489e7c2b4f1e357db769e27605462ad880
layers already present: 24; layers needed: 42 (593.2 MB)
43.56 MiB [████████████████████] (0s) Fetched layer fb35d93796e742807a8 Image contains non-ostree compatible file paths: dnf: 1 lock: 2 nvim.root: 1 systemd: 3
Wrote: ostree-unverified-registry:registry.local/bootc/lafrite-bootc@sha256:a3318fb5cd371288b77a64f3177e9b489e7c2b4f1e357db769e27605462ad880 => 03d623e3219f159d0a07623520bb6da1c8dfc2364101f238ea5f29cf4131386f

[trisse@lafrite ~]$ sudo mount -o remount,ro /sysroot

[trisse@lafrite ~]$ sudo bootc status
● Booted image: registry.local/bootc/lafrite-bootc:42
Digest: sha256:4c6b01b2bd541b1352048ef2d4f58fd9b09f9574cc9f3de533287d039d735b97 (arm64)
Version: 42.20250911.0 (2025-09-12T07:09:00Z)

Rollback image: registry.local/bootc/lafrite-bootc:42
Digest: sha256:a3318fb5cd371288b77a64f3177e9b489e7c2b4f1e357db769e27605462ad880 (arm64)
Version: 42.20250904.0 (2025-09-05T07:09:40Z)
```

Thanks in advance!

Start with /usr/lib/systemd/system/ostree-finalize-staged.service and the bootc status and upgrade paths described in the report. Reproduce a timeout on slow storage, then trace how the rollback deployment loses its base image reference. Done means the failure has an intended recovery path or no longer leaves bootc unable to upgrade, roll back, or apply another image.

Scritto dal modello di indicizzazione a partire dal testo della issue.