First off.... goals. The biggest problem in small room acoustics is the room/loudspeaker interface. When you have 15db peaks and nulls in the response it kind of dwarfs all other problems. It amazes me that so many loudspeaker designers live with it. In my opinion the first thing to do with a "cost is not an object" system is to go after the biggest problems and I'd put that at the top of my list.
Second.... I'd work on minimizing early reflections. Dr. Gedees has done some interesting research showing the audibility of <20ms early reflections has a large correlation with SQ issues. I'd be interested in a design that minimizes those effects whether it be a waveguide or other controlled dispersion technique. I'm also big on even power response so that would have to be planned into the system.
Third... low distortion, both linear and non-linear. I know many of the studies by Dr. Gedees and Dr. Klippel show that high levels of distortion are difficult to correlate with preference but I hear large (in audiophile terms) differences in low/high distortion systems even at the low levels of amplifier distortion. More expensive low distortion drivers sound better in my opinion and its not below the level of audibility. For a cost is not an object system you can afford to chase imaginary demons anyway so I'd aim for the best drivers I could source with an emphasis on low non-linear distortion at the SPL levels we are interested in achieving. I'd look mainly at odd order distortion products when choosing drivers. I'll live with higher even order garbage but not odd order products.
Fourth: I'd probably design an active system just so I can control more of the system variables. When you design a loudspeaker you are making guesses about what people are using to drive it. Since I have the Hypex UCD amplifiers available to me and they are arguably some of the best amplification you can buy I'd use them and make the entire system active.
Fifth: I've convinced myself that so-called linear phase is mainly an artifact of audiophile nervousa. In my opinion what is going on in the frequency domain is the priority so I'd design with moderately steep crossover slopes. I'd be doing it all active so a couple of op amps takes care of 4th order electrical networks. Using good design and good drivers there doesn't seem to be much benefit to going steeper unless a certain situation calls for it. This is something you don't know until you get the drivers in the system and measure. You let the measurements tell you what to do... not make up your mind before.
I don't like DSP based systems at this point. I'll admit some of my dislike is irrational as I don't have a high degree of experience in A/D & D/A conversions. You can do all kinds of neat stuff with DSPs and computers these days. My problem is that we don't know what we should be doing and what we should be leaving alone.
If you have seen the movie "The Fly" the scientist creates this teleportation device. It takes apart things at a molecular level and puts them back together on the other side. I have the same problem with DSPs as I would with that type of device.
I wouldn't get into the teleportation device and I wouldn't send my signal through a DSP unless I could convince myself it wasn't causing problems. Maybe someday I will but not at this point.
There isn't enough fundamental research about what problems we should and should not correct for me to be attracted to the room correction devices. If they can show me the fundamental research that they are basing their algorithms around I might bite but at this point they are correcting for imaginary demons while creating real ones. All the room correction I want at this point is the use of a couple bands of parametric EQ and some fiberglass. The fundamental early diffraction effects cannot be controlled electronically. Non-linear driver distortion cannot be controlled electronically (at this point) and linear phase is not worth the DSP demons.
Overall those would be my thoughts on designing a high dollar system. I've intentionally left a lot of open gaps in my thinking. There are lots of ways of implementing designs that would achieve these goals. Choices of drivers, configuration, crossovers etc.. are all choices you make once you are elbow deep in measurements, not before.