Do you really not see the benefit in avoiding the need to recreate a bunch of scaffolding around 3 separate image encoders, or is this just an excuse to dump on facebook?
It smells like a NIH project to me, but then again there might have been some very special requirements that facebook needed that was not already available in other libraries.
Not well.
At Facebook's scale, the choice of whether to resize/crop/compress images on devices or at the backend must be a complete no-brainer. Why would you stand up additional datacenters instead of just burning a little more of your user's (your "product's"?) battery?