I restored a Sony TA-8650 in the late-1990s, and owned it for about ten years. I found the preamp to be reasonably competent, but the power-amp was indeed special - smooth, clean, and powerful. I also spent some time with it on the test bench on several occasions, in an effort to understand the characteristics of the V-FET topology and its appeal.
The Sony 8650 circuit uses them in a Class B complementary source-follower output stage, and in this implementation the V-FETs' transfer function exhibits a very wide and comparatively smooth crossover region, while still achieving reasonable efficiency compared to many Class-AB designs. In the mid-1970s most higher-power bipolar amplifiers still used quasi-complementary (all-NPN) output stages, and in this context, the V-FET output stage was a quantum leap in eliminating notch distortion. The front-end of the Sony design is also very light on HF compensation, suggesting that the V-FETs were indeed much faster devices than the slow power BJTs of the era.
The major drawback is that the bias voltage requirements (Vgs for a given Ids) are extremely variable in FET production, and hence even with selection and matching, it's a very twiddly business getting the DC parameters to be consistently stable on a complementary V-FET amp. The Sony amps tended to have the compensation diodes, trimpots, and bias resistors drift around a bit, with catastrophic results of rapidly lunching the output stage. Upgrading these parts helps tremendously on older amplifiers, but it's unlikely that V-FETs could ever be manufactured to have the bias-voltage stability anywhere approaching a bipolar transistor. The Sony protection circuit also seems to have an extra network aimed at rapidly sensing an oscillation condition, which I assumed meant that they were having some issues with parasitics (very scary for a unobtanium parts!)
VFETs are also called Static Induction Transistors (SIT) & their I-V characteristics resemble that of a triode. VFETs/SITs basically behave like a voltage-controlled variable resistor (rather than a voltage controlled current source as MOSFETs & JFETs do). You already know that the triode vacuum tube is perhaps the most linear power amplification device. So, here we have a semiconductor with triode characteristics & the semiconductor, in comparison, has way more life than a vacuum tube.
No offense, but I think almost all of the above to be completely silly:
1. The circuit implementation determines the performance, and triodes aren't available in complementary versions. Thus, a complementary V-FET circuit can't be compared to a triode circuit, period. There are also a variety of reasons why a V-FET would be a poor choice for a traditional transformer-coupled triode output circuit. Maybe with some creative draftsmanship one could get the graphs to resemble each other . . .
2. The difference between a "voltage-controlled variable resistor" and a "voltage-controlled current source" is summed up wholly in the parameter of transconductance. V-FETs are indeed lower here than most modern power MOSFETS (and of course BJTs), but that doesn't in any imply some fundamental difference in their application from either. Actually, it's one of the main contributing factors to the above-mentioned bias instability.
3. I have nothing against triodes. . . but the "everybody knows that XXX" is never an argument that I can take seriously.
4. In retrospect, it seems that had V-FETs been more like triodes in the area of reliability, they might still be manufactured today. One of the most useful attributes of thermionic devices over semiconductors is their ability to endure periodic excursions beyond their normally safe operating limits . . . and here, I can't think of any audio power device more unlike a tube than a V-FET - and NOT in a good way.
To that end, one apocryphal statement worth repeating . . . never, ever, bring up a Sony V-FET amp on a Variac.