Good points. These are all the same arguments folks make in my line of work for writing source code in the clearest possible manner and letting the compiler/linker sort it out. If the code is bloated or slow, the compiler/linker needs to be improved.

In reality (the place I live, next door to theory ) a good coder can write clear code that lends itself to be well optimized by the compilers available. This also has the side effect of allowing us to create marketable products now.

Also, with the current bunch of codecs I'd be surprised if there wasn't some similarity in the data that is selected for heavy compression or discarding.

This may not map well to audio mixing, but I thought it worth mentioning. I'm not an audio engineer, and I don't even play one on TV.