Also, S2LAL illustrates that vast gains in efficiency can still be achieved for general digital computing, even in CMOS

Heh, even Apple's new chip is demonstrating that there's a lot of power slack to be picked up in "general computing" :)

Sure, but sometime in the not-too-distant future, all of the low-hanging fruit that's available in terms of architectural/packaging improvements will have been picked, and designers will run out of things they can do other than moving to energy recovery. So, this work is looking a little farther ahead, at what will be needed to make further progress.