I think what you are getting at might be this... we think of precision in a linear way, and it works for the most part, because we are not being rigorous. We are just tweaking things and seeing results directly.
However, it's no different from the way one might approach making a mathematical model in physics. First, you try a linear model, and it works but it has limitations. Then, you make a non-linear model, and it starts fitting reality better.
Remember, Newton invented The Calculus in order to be able to describe motion, because Linear Algebra was not working.
As I said, nothing about shooting is linear. That's because nothing about motion is linear.
Sure! Some things about motion are linear ("a body in motion will tend in the same motion, unless acted on by another force", etc). But you're absolutely right in that bullet drop is non linear.
Maybe the discussion has mixed accuracy and bullet drop. Accuracy depends on lots of things
that can't be calculated (a fleck of dirt in the barrel that's not there the next time, the difference in seating depth from one cartridge to another, etc). Bullet drop is very predictable
Gravity works by accelerating a body at 32ft/sec^2 . You're right in that the ^2 means the effect is non linear. The way it appears linear ( 1/2 MOA at 50 yrds =1 MOA at 100) is that the time part of that equation is REALLY tiny , especially when you square it.
The difference in time between 50 yards and 100 yards (~150 ft) at 2000 feet/sec is a small number. Square that small number and it gets REALLY small between 50 and 100 yards anyway!
distance traveled = V0 + 1/2 at^2
In the vertical dimension (bullet drop) V0 (initial vertical velocity =0) so we can ignore that
Distance= 1/2 32ft/sec^2 * t^2
Lets say it's a slow 22 and that velocity remains constant between 50 and 100 yrds (it doesnt and it's gonna get tricky calculating the negative acceleration of the bullet due to air resistance and ballistic coefficient etc. It too is non-linear, you're right. That pesky t squared!!)
Since the diff in velocity towards the target (not in "y" towards earth) between 0-50 yards and 50 -100 yards is pretty small, the error between the the two "y" dimension calculations is pretty small if we assume the same velocity in the two calculations. It's pretty close to linear here, but you're absolutely right...as the velocity gets lower the assumption gets less accurate.
Accuracy is subject to a host of unpredictable variables caused by randomness in the real world
Bullet drop and bullet impact position due to gravity and/or inaccurate aiming are very predictable (but not linear as I said). What did I goof up this time ?