There's one measurement of the O2 I've been pestering people about for years but nobody has ever bothered to conduct: Distortion vs. input level (especially high-frequency / IMD) of just the input (gain) stage, i.e. with volume turned reasonably far down and no load so that the output amps have negligible impact. The JDS Labs O2 comes with gain settings of 1X and 2.5X, if memory serves? Now I'm not sure how much there'll be to see at 2.5X, but at 6.5X one should be able to see crossover distortion coming in quite clearly, as the NJM2068 increasing struggles to drive its feedback network. (There should be no irregularities at 1X, since an opamp that couldn't drive a load consisting of nothing but a 10k pot properly at unity gain would have to be quite pathetic. Even a 4558 can do that.)
NwAvGuy himself found lower distortion and higher clipping level at 7X when using an NE5532 instead, which - surprise! - is a better load driver. Unfortunately he never tried some different feedback network resistor values - increasing those by about a factor of 2 should address this issue if you need high gains like 6.5X. I'm talking about R16/17/19 and R21-23. Ideally C19/20 values should be reduced by the same factor as well. I would also suggest doing the same or perhaps even more when building the low-power version - low power opamps invariably are not good load drivers, so why torture them needlessly? Even the TL072 (shown to be a sad loser in default configuration, being a notoriously poor load driver and preferring at least 10 kOhms or so) might perform acceptably then.
Besides, one motivation for low resistor values - aside from keeping impedance mismatch and resulting distortion low - was keeping gain stage noise down, but in practice chances are that source noise is going to swamp this anyway, so that's kind of silly. (How many sources have you seen that had less than, say, 5 µV of output noise? That's 106 dB below 1 V or 112 dB below 2 V, much like the ODAC actually. Quite respectable and more than sufficient for a source followed by another volume control, yet still almost 20 dB noisier than the suggested gain stage config at 6.5X with NJM2068. Most anything short of an LM358 - heresy! - would be adequate noise wise. Using one of those for the lulz might be worth it to those curious enough, but don't forget hacking in the obligatory output pull-down resistors to -15 V, typically 10k, which is the usual trick to avoid this part's terrible crossover distortion.)
I think this is an important test because you might have a source with a fixed output level, like a traditional CD player, and gain stage distortion solely depends on input level with nothing the user can do about it.