[ptalgulk01]

A stale cached bootc status can make the daemon re-stage and reboot an image the node already booted after a mid-rollout image change (#256). Force a live bootc status read whenever booted != desired, before any destructive action, falling back to the cached view if the refresh fails.

Adds TestStaleStatusNoSpuriousRestage (fails without the fix).

Fixes part of #256

[alicefr]

@ptalgulk01 have you been able to identify the cache as the culprit? I haven't been able to reproduce the issue in a way to have meaningful log in #251 :/. I don't know if the real reason can be really a stale cache since we reboots and read a fresh result at start-up. We should understand if is our bug or if it is bootc that reports stale data during reboots. @Johan-Liebert1 @jmarrero does this ring any bells on bootc side?

[alicefr]

We finally have a failure in the PR (see this run). From this run, it seems that the first reboot was never finalized and it got lost and it was tried again.

[Johan-Liebert1]

bootc never caches the status result. Every time bootc status is run, everything is freshly fetched from the filesystem

[alicefr]

bootc never caches the status result. Every time bootc status is run, everything is freshly fetched from the filesystem

But we do cache bootc status in the operator, but the reason seems something else