A couple of years ago there was a long and heated discussion in DIYaudio about this topic (maybe it's still going on). Generally speaking higher SNR means in the digital domain more bits which would translate into more resolution. The position from the developer (Soekris) was that the Soekris DAC would deliver 26bits of resolution (calculating it from the number of used resistors per channel in the R2R) while the other party insisted that the SNR numbers simply wouldn't verify this claim.
So whatever is true we can say at least that XLR has higher SNR compared to SE DAC.
My personal experience is that it also depends a lot on how one connects the DAC to the rest of the system and what source files (bit depth) are being used. Especially considering digital vs. analog volume control or a mix of both (like what I am doing). Because if one uses a 16bit signal and digital volume control one will listen in the end to maybe 12bits of resolution only. Can still sound satisfying with the rest of the system and room acoustics and one wouldn't even notice. With 24 bit signal depth and a 26 bit DAC (like the Soekris claims) it works as long as the input sensitivity of the power amp and the sensitivity of the speakers are in range with the necessary digital volume settings for comfortable listening.
A combination of digital and analog volume control make it easier to fine tune without losing resolution and allows better adjustment to the rest of the system in my experience.
The biggest flaw of analog volume control is that it takes some part of the signal and connects it to ground (voltage divider principle) and a lot of transients and subtle elements of the signal get lost first. The lower the settings the more is dumped to ground. A combo of digital and analog allows the best compromise between bit depth and transients. It's clearly audible, at least in my system.
So XLR or SE considering SNR, bit depth, and volume control is the chain that would need to be looked at in order to find the correct answer.