A huge design constraint in a non-ass sounding pitch shifting design is the processor power required to implement the pitch shift without varying the content length. The earliest Gadget Trojan Horsed into the Schiit Show a couple of years ago did alter program length as a function of pitch. That sort of design inherently favors non-ass of sound. The Gadget is adjustable for zero pitch shift. If it is working correctly it should then be inaudible in the chain. It is very, very close, just a RCH away, just an artifact of the S/PDIF. It took me a long time to be able to spot that. There are dozens of pitch shifters available on the net as plugins or standalones, many at no cost. They are not even in the same city, much less ballpark as their zero to zero pitch tests show. Their designs seldom, if ever allow the user realtime adjustment of the pitch mid-program to quickly find the skingasm spot.
The design problem with pitch adjustment is that a small percentage of samples need to be added or discarded. It seems there are no good sounding real time non-integer integrators or decimators that run in less than huge server sized boxes. Hence Ivana's algorithm, with the megaburrito filter at its core.
Oh, and with respect to temperament, the Gadget changes nothing. What comes in (probably linear) comes out the same. Autodetection of pitch and temperament change require more processing power and at least a Gungnir sized chassis, as well as MUCH more development time.