When accessing an entity based on a foreign entity ID, we may encounter a repository prefix that cannot be resolved; This should not happen, but it can reasonably happen, e.g. when a remote repository went offline; We then still want to be able to load (old) revisions that reference that repository, without triggering a hard error.
Essentially, EntityIds, with prefix or not, should be treated as user input, and any problems with their form and content should be handled leniently. EntityIds that refer to an unknown repository should be treated like EntityIds that refer to a non-existign local item. If possible (e.g. when throwing an exception), additional information should be given. But in situations where there is no back-channel for such information, it's OK for IDs refering to an unknown repo to behave just like IDs refering to an unknown entity on a known repo.