It all depends on the average value.
My experience pulling commercial and military loads is that you find some fairly wide charge weight variation. Larger than home loading equipment would make. The .30-06 M2 ball lot I pulled down 20 rounds of was loaded with WC852 and had a span of 0.8 grains. This is remarkable because I don't own even a cheap powder measure or hand scoops that can throw a spherical powder that inconsistently, but Lake City's high speed loading equipment managed to do it. The Federal Gold Medal Match .308 Winchester with 168 grain SMK I've pulled had a 0.4 grain span of IMR4064. I had one lots of foreign-made 7.62 NATO ammunition about twenty years ago (Portuguese, maybe; I don't recall clearly) years ago that had a charge weight span of 2.0 grains over 20 pulled rounds.
But here's the thing. If you have a load that is a sloppy ±1.0 grain precision, but your average load is at least 1.0 grain below maximum, that's not unsafe in any way. And sometimes it can still be pretty accurate. Hatcher Notebook includes his mention of having one year loaded National Match ammunition with a coarse grain powder that the arsenal's loading equipment could only dispense to a span precision of 1.7 grains. He'd compared it to a shorter grain candidate that the arsenal equipment could dispense with a span precision of 0.6 grains, but the wider span powder's loads were consistently more accurate in their testing. So that powder wound up loaded in that year's National Match ammunition and several records were set with it.
Hatcher attributed this performance to the coarse powder's ignition characteristics. Packed more tightly and with the grains closer together it lit up more slowly and acted like a slower burning powder, thus compensating for the greater weight of the charge. The result was apparently a sort of self-regulation of performance over the load range they were in.
You will see the above phenomena mentioned from time to time. It's usually in the form of an argument that it's better to dispense powder by volume, despite having to tolerate some weight variation, than to dispense by weight. This is precisely because only volume dispensing takes round-to-round differences in charge packing density into account.
In practice, though, most powders work pretty well when dispensed either way. What you want to do with accuracy loads is find one that's relatively insensitive to exact charge. That is, one that you can intentionally vary ± a third of a grain or more and not have the point of impact on the target change significantly. Then you can adjust your powder dispensing gear to average at the center of that range, and figure both weight and charge density can then vary a little without affecting the size of the groups you can shoot with. At that point you can dispense it by either method with equal performance expectations.
Board member
Dan Newberry has a system for finding such tolerant loads intentionally.