The THAT chipsets are a very nice starting point. I have designed a few compressors over the years and never had such neat connect-the-dots solutions.
I guess this depends on how much time and effort you are willing to invest in making something truly different than the other hundred canned solutions out there, based on the same parts and same app notes.
I would suggest a two PCB solution, where one circuit board handles the basic I/O and VCA block. With a second circuit board to handle all the side chain manipulations. This side chain processing is where the rubber meets the road and where you will spend all your time tweaking stuff to make subtle changes. Note: One VCA path could be made into multiple different products by attaching different side chain processing (gates, de-esser, etc).
I see several useful circuit tricks in your cut and paste schematic that I have used before, but the design trade offs in even one small section would require a lot of words and time to describe what it is going on, let alone how to optimize it (as if there is a single optimal solution).
Compression has some general "always good" and "always bad" characteristics, and other less objective fashion elements. In other words there can be more than one good compressor solution. In an ideal world you can cover several "good" compressors with one and a few controls, but the world is rarely ideal.
I have wasted more time than I should thinking about an ultimate analog dynamics engine, and I have pretty much settled on a digital side chain, and computer interface to handle the too many knobs, switches, and variables I want to control, while still keeping it simple to use. I still like the analog VCA path, for now (while I have some ideas about possible improvement there too that involve trade offs. Everything involves trade-offs).
Unfortunately I neither have all the tools to design the full computer interface, nor do I think my design time is well spent on making yet another fancy compressor, when there are still green fields to harvest that haven't been this fully explored.
JR