And I'm leaning that way not so much for efficiency, but for running smoother on startup. Maybe it's the fairly narrow magnetic poles inherent in the Neu design, but it seems that running it higher rpm and gearing down made for a much smoother start. Rough, almost-cogging starts probably pulls much more current than smooth starts, so there's some efficiency lost there I'm sure.
It probably also matters how you are using the motor. If running a technical track where non-WOT fine-control is needed, this may be better. If running WOT a lot on a track, you probably don't want to wind it out. Then again, if you are going for pure high speed runs, efficiency doesn't really matter.
I've run the 1512 2d (2600kv) in my Hyper8 at close to 20v (16 NiMH cells - just temporary for testing) on my buggy and it seems to like it. Temps were very close to using 4s lipo, so efficiency seems about on par. Granted, I don't have an eagletree to confirm this, so I'm just using the scientific runtime vs smoothness vs heat method.

It is possible that at higher rpms, the Neu may WANT to be geared a little higher. I know my 16 NiMH cells are definitely heavier than 4s lipo - the added weight can be compared to gearing up a little.
Even when I ran my 4s setup geared 46/16, it was still fairly cool, yet stupid fast for a small track. I did notice that geared that high, the starts weren't quite as smooth - acted like it was on the edge of cogging. When moving to a higher voltage (16 cells) and gearing down, the starts were definitely smoother despite the added weight. I was really looking for smoothness for a small track, so I didn't try that at constant high speed.
I recently tried my 151
5 2.5d in the buggy on 20 cells, geared 46/16 and got about the same temps at the 1512 2d on 4s geared 46/14. But the torque was too much. I don't know if the temps were because I couldn't get traction or what. But the starts weren't quite as smooth - I just figured it was because the ESC wasn't getting enough rpms at slow speeds to effectively commutate.
I was kinda limited to the Quark's voltage threshold for my tests. If I can get some time, I might try my HV-110 instead to open up more options. Then, I can try 4s, 6s, 8s on my two motors with different gearing. Then put together some type of chart to get temps vs speed vs current draw vs efficiency, etc.
I'm sure it matters heavily on which ESC is being used too. The CC ESC startup routine certainly acts much different than the Quarks. Even under stall, the Quarks seem much more controlled. However, that was before the MM with Neu update or the MMM. I haven't run a Neu on a MM since the initial test, so maybe it's not the same anymore.