Velocity is the key issue because that combined with the twist rate is what gives the resultant RPM on the projectile. RPM's that are too-high on a round ball are just as detrimental to accuracy as RPM's that are too-low on a conical bullet.
The purpose of applying the twist to round balls is to correct for the minor imperfections in the projectile. Once you exceed the RPM's required to "correct" for the flaws, the flaws will then become amplified. Ex: When tires on your car are spin balanced, they are done so at 45-65 Mph depending on the particular machine. Normally at speeds below 45 Mph, there isn't enough balance loss to be noticeable because the tire & wheel assembly cannot develop enough mass at the given rotational velocity to be felt through the vehicle suspension. If you exceed the upper limit by say 15 Mph putting you at 80 Mph, you may start to feel some vibration because now the tire & wheel is developing enough mass from the velocity so that the vibrations are no longer absorbed by the suspension.
The same with a projectile - within a given RPM range, the projectile will be relatively balanced and fly straight. High quality round balls that aren't deformed upon loading or firing require very little spin to keep them flying straight however, when you put too-much rotational force on the ball, the minor flaws become amplified and the flight path begins to look like a corkscrew because the heavy side will always be pulling on the light side. This is why with a given load you can get a one ragged hole group at 25yds but at 50yds the group will be opened considerably - but, if you reduce the powder charge thus reducing the MV (muzzle velocity) which reduces the RPM's the 50yd group will tighten up.
Bullet RPM = MV X 720/Twist Rate (in inches)
Using the above formula, a PRB leaving a 1:48 twist bore at 1600fps will be spinning at 24,000rpm. The same ball leaving the same 1:48 bore at 2,400fps is turning at 36,000rpm.
All things must be taken into consideration too. If you're shooting a soft lead ball combined with a very tight patch, when you're beating it into the bore the ball is no longer going to be a spherical shape. The quality of the ball before it's loaded will also make a difference in that voids or contamination within the ball, a large sprue, ball that's not perfectly round, has flat spots or any other mechanical issue will affect accuracy.
The amount of variables make it impossible to identify a "optimum" twist as the optimum twist will be determined by the variables applied by the user. Someone running dropped buckshot or whatever balls that fall from the mold will require more RPM's on the ball for a given load than someone who is running very high quality balls. Same as someone running very light powder charges producing lower MV can get away with a faster twist than someone wanting to run the ball at higher velocity with larger powder charges. Tight control of the variables are what allows some folks to run smoothbores that are exceptionally accurate with PRB's despite the fact there is no twist at all applied to the ball.