I have designed and built several new single band HF rigs for my own use since the R2pro, but have not done any complete transceiver designs intended for others to duplicate. The R2pro was specifically designed as a set of modules so that a designer can combine an assortment of those modules with others and come up with a transceiver optimized for a particular, personal application. Much of my recent most design work has been for instrumentation at VHF and UHF, so it doesn't really lend itself to the "high performance at HF" realm. The R2pro design still holds up well in 2016.
Regarding being inspired to design a new high performance transceiver, I find complete transceiver designs uninspiring, hi. I personally enjoy coming up with a particular application that isn't well served by anything currently available, and working on some unique design for that particular niche. Occasionally it has been a large enough niche that I've worked with someone like Bill Kelsey to come up with a commercially available kit of parts, but more often I just design and build a few to fill the immediate need in my lab. Sometimes that "need" is just a whim, and some of my most enjoyable and technically rewarding work has been to satisfy some personal curiosity or other.
Regarding your wish list:
"Best Sounding Audio Possible" is still the same as it has been since my first receiver with diode ring mixers to do the frequency conversion before any gain, followed by a serious audio amplifier string drawn from the audio recording industry. In most of my recent work, starting with the R2pro, the limitation on dynamic range inside the audio channel is set by the op-amps. About 80 dB from the noise floor to non-harmonic artifacts is relatively easy, but you have to be careful. If the noise floor at your receiver volume control is 80 dB below a volt, that's about 100uV, and well below what you can see on an oscilloscope. These are very old principles. Though they haven't been common in the communications receiver field, no breakthroughs are needed, just careful design. The R2pro is better than it needs to be.
"Good Dynamic Range" is an interesting concept. I like to design receivers that have good enough dynamic range, which is different for every application, on every band. Unlike during the 1970s, today most of our receivers have enough dynamic range, in the same sense that cars have had enough wheels since designers figured out that 3 was not enough. Once you have enough, adding more is not an improvement.
Some of my recent receivers have had a crystal filter between the antenna and first active stage. Those are stellar, but that's not usually needed. They do illustrate the point that if you find yourself operating in a hostile EM environment with many very strong nearby signals, there are other ways to address the problem besides in the receiver circuitry. Ever since I achieved "good enough" for all my personal applications a while back, I've focused on other receiver performance measures. In particular I try to avoid sacrificing something important to make an improvement in a specification that is already good enough. I have never been on a hilltop with Wes Hayward when anyone missed a contact because of inadequate dynamic range, but I have been when we missed contacts because the batteries died.
"SCAF type filtering." I've played with switched capacitor audio filters and other commutating signal processing since the mid 1970s. Each time I've built one into a receiver I've run into problems with that danged clock. I tend to listen at and below the noise floor, and one of my pet peeves is when I can hear anything at all other than pure thermal noise when the receiver is terminated with a room temperature 50 ohm resistor. Since I routinely record signals for post processing in a DSP system, it is disheartening to discover that some harmonic of the SCAF clock beat against the LO at a particular frequency, ten dB below the noise so you didn't hear it when you made the recording, but 10 dB above the noise floor in your FFT waterfall plot. So I have avoided such things for my last few decades of instrumentation-grade receiver designs. When I have a digital dial, I include a switch so it can be turned off when I'm recording weak signals. None of this applies to you--please continue to experiment with SCAF filters, embedded processors and other generators of digital noise that may well be far beneath the threshold of signals and antenna noise for your particular application.
"Audio output power." In my lab I generally find about a half watt is more than enough for a radio tuned to a ham band. In other applications I go as high as 40 watts of audio at the threshold of detectable two-tone IM products. All of that is nicely covered in the audio literature, and there are a number of good designs available on the web. The R2pro audio amplifier in EMRFD works well driving a compact, efficient speaker in a quiet room. For more power output, use bigger transistors and raise the supply voltage up to about 24. Above that, you need to add Darlington drivers, as in the original R1 and R2 circuit, and you can then raise the voltage up to around 40v with NE5532 op-amps and get more than 20 watts of superb clean audio. That R2pro audio output stage in EMRFD has been borrowed and modified for more than one very high end professional audio application.
"RF Power" For either audio or RF power, I encourage experimenters to use higher voltages than 12. Within a few years we should have available some nice GaN transistors that will allow us to use supply voltages up around 80, which really makes it easy to generate many watts of power into either 50 ohm or 8 ohm loads. If you don't want to wait for GaN, there are some huge N channel depletion mode FETs that run the electrons in vacuum and will easily handle many hundreds of volts on the drain. They need a separate low voltage supply at an amp or so to generate the electron stream, but are capable of astounding performance, particularly in narrow band applications. If they hadn't been invented 100 years ago, we'd be all excited about the possibilities now. A 6146 would just idle along at 5-10w out and last forever.
For my most recent contacts on 40m CW, I used an R2pro and ran a very stable premixed JFET Hartley VFO driving a 6C4 and 6AQ5 output stage.
Enjoy the experiments.