If ball-to-ball friction produces residue, that would effect the cups too.
Seems to me that peak load occurs on one ball at a time, and that having the balls closer together does not reduce peak load, but reduces the time that load is applied. If degradation phenomena such as micro-welding are primarily a function of load rather than time, cup degradation should be unaffected by relatively small differences in ball count.
I can tell you that a headset with loose balls will most definitely last longer than a headset with caged balls, all other things being equal. Headsets are different than hubs/bottom brackets, yes. But it's interesting that on older, quality bikes, before cartridge bearing bb's and the like, the places that caged balls were commonly used are also the most difficult and time consuming places to pack loose balls: the headset and bottom bracket. I think it was to reduce assembly time, nothing more. Quality hubs always used loose balls.
Now with these "higher tech" cages that are being used with Dura Ace hubs, etc., it seems likely that the benefits you mention can be put to use without the drawbacks of cheap retainers. To be clear, I have no scientific evidence of this of course, it's all speculation on my part, but after packing a lot of ball bearing assemblies over the years on bicycles, there's no way I'll be convinced that an assembly with a cheap "tin" cage as a retainer will spin with less friction than an assembly with loose balls. I'll concede that the durability issue may be a wash with something like a cup and cone bottom bracket, because a quality bottom bracket has very durable cups, likely to last a very long time regardless. But you can see, feel, and even hear the difference in terms of smooth operation-