Comment 7 for bug 261315

We also need to decide if the fallback repositories should be seen at the RemoteRepository level, or only inside the _real_repository, or both. It seems to me like the first is better; the last might cause confusion with the repository being searched twice.

I'm kind of surprised there is not already a test failing for this: we should have a per-repository test for stacking, and the scenarios ought to be configuring that to be accessed across a remote repository.

We don't seem to actually be calling RemoteBranch._ensure_real, so my first attempt at configuring the fallbacks using the _real_branch isn't working.

Hooking this in at the creation of the RemoteBranch does seem to work so now I'm going to see why it's not tested, or how to test it.