Nick, all the classic soundstage expansion cues: better height resolution, then adding in significant more depth resolution, more relaxed presentation. Absolutely delightful (and a delightful surprise to know that there is even more SQ to reveal...mScaler is the gift that keeps on giving).
I have been hearing similar things with my end point experiments: NUC running real time OS (Audio Linux) in memory with Squeezelite with extremely large memory buffers on the network side and the ALSA driver side. The SQ lift from these experiments (significant) is very similar to what I hear with the ISORegen/LPS 1.2 combination. In fact, by chance I had 2 ISO Regens in house, and when I put them in series, it is the closest I've come to not hearing a difference with the end point optimizations I've been doing (which is awesome).
Clean power on Vbus to HMS has great intuitive appeal to me (LPS 1.2 on the ISO Regen), and I get squint my eyes and say "yeah, that makes sense that it makes a difference". The only hypothesis I have for how the hell real time end point optimizations or multiple USB signal regeneration steps can be so audible is some sort of signal integrity induced noise on the USB receiver side. If ever there was a clutching at straws hypothesis, this is it. Much more thinking and experimenting to do to see what is really going on, but the pragmatic result is awesome.
In the interim, Alex and team are extremely generous with their home trial set up (pay for shipping, try it for 30 days). If others try ISORegen and LPS 1.2 with their mScaler setups, please share your experiences. I don't know how much of what I'm hearing is unique to my setup and house, and what may be tickling at a more general noise vector. I look forward to hearing what other people are hearing.