I'm curious on this. Why wouldn't statistical compression go smaller? Assuming you restrict the statistics in question to be tailored to the metadata that is being compressed, it feels like it should be comparable? Yes, you may need/want to store a precomputed dictionary at the receiving end so that you aren't sending that every time, but that is really no different from having custom code that you need at receiving end. Right?
Yes, if we can. The users of Fury may be able to do this. Fury may only provide such an interface to users to let them pass such an dict/zstd/huffman or something like