The vaguely logarithmic trend of motor life tripling for each volt less may be over.
Something went catastrophically wrong with the first motor I tested at 4V, its speed started slowly dropping at around eight hours, and it died right around 14 hours (an hour and a half LESS than the 5V motor). I tested another motor at 4V, which has a failure curve that looks much more like the previous tests, but it only lasted about 17 and a half hours, just two hours longer than the 5V motor.
The stats so far:
My guess is that at higher voltages, arcing is the primary source of brush and commutator damage, but at these lower voltages and longer lifetimes heat and friction are limiting the time and cycle life as well. The motors all look the same inside in the end, but at lower voltages the motors also seem to be venting smoke, which builds up as soot on the windings and motor cases. It may be that this happens all the time, but at higher voltages there is less time for the soot to build up.