I don't know the technical details either (not a EE), but the prevailing opinion is that J-FETs sound more "natural" and have "warmth" compared to BJTs. That seems to be what Larry prefers in sound. Whether you believe opamps have a sound or not is not the question here --- if you want to choose an opamp based on spec, for a transimpedance circuit for I/V conversion purposes, the majority of suggestions that I've seen have been for BJT parts, but in many designs, even very high-end designs, I see a lot of people putting in J-FET. I guess the key is that if it doesn't perform any worse, why not go for the part that "sounds better"?
I believe the reason was that it requires less power but sounds similar to it's big brethren. Some specs might be slightly lower but for all intents and purposes it's the same.
His reason was, with the proper tweaks, ES9018K2M sounds better than the desktop model.
However, the Wave comes on a different model, it's a ES9018K2MV. Asking Larry the difference, he said, he can apply more magic.
Another worthy info is that this "V' model is the same chip used by the Pono.
It's the same IC, it's just that a few features are unlocked and "overt", as in documented in the NDA version of their datasheet. So you can now more easily bypass certain settings, such as the turning off the integrated 8x interpolation filter (and putting your own in, like the DAD perks). You can now also selectively turn different filters on, from the standard linear phase FIR filter, a minimum phase FIR, IIR filter, etc. along with DPLL timing, etc.
Basically, ESS want to control how much information gets out into the open, and that's why, historically, DIYers don't like using ESS chips, because they have no idea how to use them as the publicly available datasheets basically said nothing. ESS wants to know who's using their chips and where. However, the K2M chip is a higher-volume chip that was partly made in response to strong demand from smartphone makers in China. The upcoming SABRE9601 is also a part made "for the Chinese". Now that large volume clients are using their chips, they need to make more options more accessible to engineers.
For the ES9018S, a lot of functions weren't switchable by default, you had to "trick" it into certain modes, which was more work for the DAC designer. That's likely the big reason why people are favoring the ES9018K2M --- it's just easier to design around because the option for "on or off" is right there for almost all features.
LH should also be an "official" client, as Larry mentioned somewhere that the design heads of ESS (Dustin Forman and Martin Mallinson) were involved in technical support for the GW. R&D for ESSTech is located in Kelowna, BC, which is a stone's throw from Sacramento (okay, more like a skip and a hop), but the point is that they're in the same time zone, so Larry can confer with the ESS guys in real time.