Just to make my life extra fun, it seems Fedora 24 has a third base location. Nothing else exists in that alt/unofficial tree, only four Fedora 24 images (SoaS and Design Suite for x86_64 and i686).

IIRC the reason for this is we still had #1315541 - AKA the live image compose lottery bug - going on when Fedora 24 was released, and those two images lost the lottery for the final compose. Since they're well-tended spins and their authors were sad about them not being in the compose, Dennis grabbed the images for them from a late nightly compose and made up that location to hold them.

It would be nice, I guess, for fedfind to find them; though technically they never were part of the real Fedora 24 Final compose (they come from an entirely different compose), I'd say the fact that we link to them from e.g. here without any kind of note that they're not 'real' Fedora 24 images trumps that and fedfind should just treat them as part of the Fedora 24 image set.

Unfortunately it'll be hard to do this at all transparently with fedfind's design. My current idea is to add a helper for creating image URLs (since even just for 'regular' mirrored composes it's now not entirely trivial) and rejig the 'base locations' a bit, maybe have a dict whose keys are the 'names' of the locations and values are the actual locations, and the image dicts would have a location_type key or something whose value would be the location 'name' from the locations dict, thus you can easily look up which base location to use for which image...that seems a bit nicer than just inventing a third property besides location and alt_location, and running the risk of having to invent a fourth in future if any more crazy shenanigans happen...