Very impressive engineering work.
Their custom-manufactured hardware must have taken years to develop. All on a non-commercial university low-budget. Energy saving gains are very impressive.
This brings back fond memories of failures and (small) fires. (23 years ago I wrote the early Linux device driver for dynamic voltage scaling)
I never understood why we pour clock signal when that signal should be budgeted. Like, I know some computation needs x clock cycles to be done. Why not provide only ~x clock cycles to the component?
A lot of (low level) code is written with the assumption that the clock rate is fixed. Sometimes bus protocols like i2c are implemented with "bitbanging" where the code "waits" for the right amount of time by doing a bunch of nops in order, assuming that each nop takes the same amount of time.