Air resistance has a significant impact on a falling Phantom with the shape and the propellers providing drag.
It doesn't continue to accelerate but reaches a terminal velocity that varies between 28-35 mph.
GPS shouldn't have much to do with it as the GPS is only used to provide horizontal position.
8 to 35mph for terminal velocity doesn't seem much to me as the phantom is already able of this speed horizontally... I don't know how you calculated this but I would be interested to know. Even if the drag was 50% wich is definitely not the case (I would estimate it at max 20%) the fall should be much faster.
Anyway, even at that speed, if the motors are stopped 5 sec and restart, they will need at least another 2 sec for the CSC to start and maybe another 5 to reach the full throttle. In just 7sec, the phantom will have fallen more than 100m before starting its slowing down and recovery.
Where the GPS mode has a major role is to balance the quad horizontally, and stop crazy drifts after the manoeuver. This is the only way to stabilize it, so yes it would be essential in that case. The problem is that it doesn't give the full power available to the engines, and I wonder how long it would really take to stop the fall efficiently.
Now I can say that I have kind of experienced something not similar but near I guess. I have made some 360 flips over in manual with my P2

, he he... going very high, idle the throttle and roll max to the right. At that very moment, the phantom is almost stil, and start a slow flip over. this moves loses probably about 7/8 meters with my stock gains.
Swiching back to GPS when the quad is at 90 degrees brings it to horizontal but looses quite a lot of altitude. probably at leat 20m
Switching back to GPS at the end of the looping and it falls like a stone for I estimate at least 30m, and another 10 meters when you hear the engines reving like mad, but late and not that fast anyway.
Throtteling manual again at the end of the looping saves the day. with a minimal loss of altitude. would say about 5 m.
My point is: I don't think that the phantom can recover from a stop and start in the air... I wouldn't try that.