I tend to think the SMSL design - even though it is almost exactly the datasheet example circuit is a bit odd.
As noted, they have ignored one half of the DAC. There are two channels of output that are not used. The Topping design uses all 4 channels and sums pairs. This should lead to less noise and generally a small increment in accuracy.
The AKM datasheet depicts summing all 4 outputs for a mono DAC. I can't see why a designer would not go for summing pairs. It is a few cents on the BOM. (Maybe AKM offer a secret cheap part when one half doesn't meet spec
)
Output buffer is a curious question. I had not twigged that the 100Ω and 510Ω resistors were part of the output filter. They are mixed in with the mute transistors on the PCB. In principle an output impedance of about 1kΩ is no big deal. It simply forms a potential divider wit the input impedance of whatever it feeds. So feeding a 10kΩ input means you get 91% of the output voltage going into your device. A 100Ω output impedance would see 99%. Where it might start to be noticeable is feeding devices with unusually low input impedances. In the extreme, say 1kΩ input, you get only 50% of the input voltage.
If you used cables with noticeable capacitance you might get a just perceivable roll off in the high end. Certainly there was evidence that the golden eared could use different cables as tone controls, and ascribe the subtle change in sound to the special qualities of the unobtainium eye-wateringly expensive design.
But connecting the output filter, unbuffered, straight to the outside world does bother me. Maybe it is fine. It just bothers me. I can't help but feel that AKM were stretching for the best possible measurements on a reference design.
IMHO the Topping is a much nicer design all round. OTOH, I am more than happy with DX3 Pro and D50s. So I'm not going to be springing for any of these ultra high performance DACs. We have already reached peak DAC.