How many samples of drift should we really anticipate between even two crappy USB DACs? I
tried to find some measurement of drift and couldn't even find posts about it here or elsewhere.
What is the worst case scenario for de-synchronization between two DACs? I don't really know, but the OS is presumably sending frames to the DACs at the exact same time. A packet is 125us long (0.125ms) - so if one DAC gets a full frame behind that's how far off you get. I'm comfortable calling that discrepancy irrelevant when the use case is to delay sound by 25ms for sub integration.
For tweeter integration we might be very concerned about this amount of delay, but here it's probably not worth worrying about even if variable.
If there are disruptions it could be buffer underruns, you may need to increase the buffer length. I have also read similar comments about Voicemeeter, but have been using it in my main system without problems (well, without problems caused by Voicemeeter) 24/7 for some months now... perhaps they've fixed the bugs.
Anyway, the tutorial I linked earlier specifies either Voicemeeter Banana, or Voicemeeter Potato, I guess these are forks of the original, which support multiple device output.
If the windows-based approach works, I think it will only cost you $30-50, worth a try IMO.
This one is $55 but I bet there are cheaper ones.