We lack a proper and consistent mechanism to detect and handle a known origin that has become unavailable (typically a deleted github repo).
For now, this generates tracebacks (eg. https://sentry.softwareheritage.org/organizations/swh/issues/1786) which is not great.
The problem being that each origin provider (forges) will have its how behavior so it's most probable that a simple solution is not that easy to come with.
Also, as @olasd suggested on IRC:
> <+olasd> we also probably need a "origin not found" origin visit state
> <+olasd> (that's distinct from a plain failure)