Blast from the past for me, though primarily interacted with the complementary Whonix side of things. Not surprising to read, considering how lean Qubes was from the get-go designed to be it makes sense that most things are from resulting upstream rather than with their code.

Fully aware that it was never the goal for Qubes, but I have never been able to shake the idea that one could leverage their architecture in ways beyond security hardening, especially that screenshot with MSFT Office running in its own guest got my mind spinning back then. Might be worth revisiting some old ideas I'm just recalling, especially with there having been over a decade in development across many projects focused on hypervisors by many smart people, making a few old experiments likely less impossible.

Author of the paper here; AMA.