Comment 4 for bug 1233924

So the code in question *does* have multiple search locations, but we ended up with a bit of split-brain.

The highest level loop says:
  for source in sources:
     try to get signed data
     if there was a problem:
       try to get unsigned data
    if there wasn't an error:
      break
Which is, essentially, reach each location until you don't get an error.

However for Azure one of the locations that was injected was the "daily" stream. However if you aren't using the daily images, then there are no "released" images in that stream.

The currently logic is designed such that once we find a valid simplestreams we stop looking at fallbacks. This is possibly valid in the case of users uploading their own custom images. (eg I want you to use my special image, and just fail if you can't) vs (eg I want you to use my custom images for precise-amd64, but use the generic ones for saucy-i386)

Right now, injecting a valid stream into the search path is intentionally preventing fallbacks from being used.
However, we can just remove the daily stream from the Azure search path. Users that really need it can set image-stream: daily and imagemetadata-url: http://cloud-images.ubuntu.com/daily and it should still work for them.