Looks very nice - my only question is why put a tantalum capacitor on audio path instead of general purpose aluminium electrolytic?
You also might want to check which one is cheaper - bunch of 0.1% resistors or one R2R ladder component for DAC applications with guaranteed 0.5 LSB error.
I also built my CVX-3 a while ago, with accompanying 100k 0.1% resistors.
I used 29 of the resistors when building and measured all of them after soldering with a Keithley 2000 multimeter that has not been calibrated for over 10 years.
The average of resistances were 99.965 kohms so either the resistors were just biased towards low end or it could just be the multimeter.
All resistors were within 0.071% of the ideal 100 kohms, or within 0.055% of the measured average.
Four of the resistors were paralleled to make the 25kohm output resistance to ground like in original Covox.
I did not use any filter caps, and bypassed the AC coupling cap (just because I had no suitable electrolytics lying around).
To test I used a game with Covox support (it uses ModPlay unit by Mark J. Cox to play music at max 22kHz).
At first I used very long cable to Cambridge Soundworks 2.1 amplified speakers, and it sounded great at first, but I think it is lacking high frequency content.
Next I tried to record the output with another PC, just by connecting with short cable to line input on motherboard-integrated sound hardware, and this time I think there is too much high frequency content, and I think the PC line input may have some issues with the DC coupled signal.
So clearly the sound quality depends on the cable and receiving equipment due to the large output impedance of the Covox.
On a side note, I know the PC currently driving the Covox is almost too modern - it has 3.3V CMOS output drivers on the parallel port so it can drive huge loads symmetrically.