Well, some while back I did the math for what radius curves would give fully laminar flow at typical intake velocities, and it was around 4 inches, so optimal would be a bit far off what fits in an intake...
Also, as I alluded to above, what works for one, may not work for many, the flow is across them, they're not in free air, and they're not being fed from in front, so #1 gets to breathe good, but look at the cross section of the intake path from the POV of #2 #3 #4, there's all sorts of junk in the way, flow is gonna be all messed up and significantly slower at #4 than #1... so while you get the theoretical best velocity at #1, the rest might not be close. The only argument for it is the runner length increase, but you could have that and a flat sided tube that wasn't as turbulent or obstructed for the other cylinders..
Anyhoo, I suspect useful radii to start at about 1/4" and I suspect that a "3 angle valve job" radii would get a large chunk over square end, each refinement to a perfect curve gaining less and less as you go on.
IF you've got one intake, one box, such as in an airfilter box, sure do the raised one, even a square end pipe does better raised off the back surface there, but that's when you've got end on flow and only one of them...
IMO 3% flow improvement into each cylinder with a quickly assembled eyeball ground radius on the entries would be better than the complex arrangement where it's likely to go 4% 3% 2% 1% because of the side on profile of the trumpets disrupting the airflow to subsequent cylinders.
Also, I bet there's a lot of intakes you could achieve this effect on with a flexible bar hone, going up from the motor side and letting it splay at the top of the runner. You'd have to peer down the intake side with a flashlight to see what you're doing and it would take some finesse with tool speed and position, but you could get the edges off.