Just an opinion, but 223 is sensitive to brass prep and charge weights once you go out to mid range or beyond.
The ability to stay in tune depends on the composite of many variables that all have dispersion and in addition, the tuning concept isn't linear.
Teasing out the amount due to one thing like charge weight isn't easy, but to get a perspective without doing a bunch of statistics means you have to just look at the slope of the charge versus velocity plot.
Once you are willing to say we all know the point of impact and the tune are not perfectly linear with speed, then we can go forward with a simple look at the vertical change due to velocity at some distance.
Say for example a 80.5 grain Berger bullet with a typical average speed of 2750.
It will have a drop of about 15 MOA at 600 yards.
Now if you assume something like Varget gives roughly 100 fps per grain, and it takes about a change from 2750 to 2675 to cause the drop to go to 16 MOA, then the math is roughly 0.75 grains to cause the shift.
So, if everything else was held to zero, the assumption implies the charge would be able to have a +/- 0.375 grain tolerance to hold 1 MOA at 600 yards.
However, since everything else is not held to the assumption and the reality is that tuning is non-linear, the typical charge tolerance is much smaller than +/- 0.375 grains, just to hold 1 MOA.
Also, since many folks here are expecting performance much better than 1 MOA, and in many instances much smaller than 1/2 MOA, then we must also assume that the powder charge and velocity changes must be held to less than half the above in our thought experiment.
This means an ES of less than 75 fps/2 = 37.5 fps for a linear assumption, which we would all agree isn't good enough when we have to give some room for other errors.
That just cut the +/- 0.375 grains in less than half, so +/- 0.1875 if everything else was perfect and tuning was linear, and much less than that for the real world since tuning isn't linear and everything else isn't perfect either.
So to estimate a charge tolerance, just look at the slope of the powder charge versus velocity change for a given example, and then plug the velocity change into a ballistic program for the amount of drop change you would accept. The reality is usually much smaller than this assumption. YMMV