That's an awesome improvement, but I doubt it'll make it into the 2.0 -- it's memory is pretty much maxed out in v1.6. The 2.1 boards, however, have more memory . . .
I imagine it would require user to flip a mode switch into "safe mode". The tests were probblay run asssuming the rotor was goign to "randomly" fail so I doubt the algorythm would play nice with typical or agressive flight, but if I've got a quad 100' in the air and something goes wrong, I still have some time to switch modes and slow down before it lawn-darts.