*Puts on nerd hat*
Ball speed will be a function of many things, including but not limited to era, coil/flipper strength, flipper linkage design, pitch of PF, wax...etc.
Let's say Stern has the "snappiest" (strongest) flippers, in theory the fastest a ball could go would be at the very end of a flip, just as the ball leaves contact with the flipper rubber, on a low pitched game, freshly waxed, and probably was screaming down the inlane to carry as much momentum as possible into the flip. As soon as it leaves the flipper tip, it's a ball rolling up hill and will instantly start to lose speed, even if not observable by the naked eye.
Another issue with calculating up to MPH is that if we wanted to measure the speed at this point off the flipper tip, any error in measurement over these few inches gets multiplied up to miles. Same thing with milliseconds being blown up to hours.
Lastly, a single point of reference (i.e. a camera) over a grid of lines can introduce error based on distance from said grid and resulting angles. The closer the camera, and the wider the distance being measured, the more of an angle exists between the camera, start, and finish. This means you may not log the exact moment something crosses a plane, as the camera is not aligned to that plane. This is why the Mythbusters always bring the highspeed camera far away from the grid when calculating speed. Perhaps photogates would be a good way to measure something this fast in a modern Stern?
TLDR: It's a super cool test, and I'd estimate newer games are producing speeds far in excess of 3MPH... but with the potential equipment, measurement, and calculation errors involved, to button down exact speeds would require a fair amount of work and control to ensure accuracy.