The torque does drop after the threshold (as it does with any motor), and with a brushed DC motor it drops rapidly and all of the above applies. With an induction motor or PM AC motor the torque tends to drop in proportion to speed so power tends to be reasonably constant for a substantial range of speed above the threshold.
Thanks - excellent illustration
Dropping torque doesn't necessarily imply dropping efficiency. Anyone who wants to put in the effort can compare the total mechanical power (the sum of front and rear power; power for each axle is the product of torque and speed). The "power" curve on the graph must be electrical power, since it is not zero in the brake-stand stall leading up to the launch (at zero on the time scale); that means that any change in efficiency could be calculated (but not absolute efficiency, because the actual power is not known, unless you know the tire rolling radius).
I found the discussion in thread in which this graph was posted (Chassis CAN Logging To ASCII Text Plus Graphing, page 9), and the data appears to be all from the car's internal network, collected by CAN messages... so there's likely no real torque measurement at all.
I wondered about the validity of the torque data, since some dynamometer methods are questionable, (although I eventually realized that there's no dyno here) and in this case the torque values are inconsistent: from 415 somethings @ 35 mph to 190 @ 70 mph would make sense for the front because it would be about a 10% power drop, but from 200 somethings @ 35 mph to 115 @ 70 mph would be about a 30% power increase. If the real mechanical power is almost constant, the almost constant electrical power draw suggests that efficiency is roughly unchanged.
I think the torque data might be a little flaky, since the front does not drop enough with the speed increase to be plausible with constant power, although the car could be shifting power to the front with increasing speed, which makes physical sense. The electrical power data is at least plausible, because to accelerate the two-ton mass accelerating at 10 m/s2 through 50 km/h or 30 mph takes 280 kW, the car weighs somewhat more than two tons, and there is rolling and aero drag... so the 400+ kW of power consumption may be a little high but at least is in the right ballpark.
I can see why Tesla Motors gave up on multi-ratio transmissions after their first attempt failed (in the Roadster). They're now putting so much power in these cars that the speed-dependent power under the threshold (which corresponds with 30 mph in this test) doesn't matter: the constant torque at and below the threshold is probably all the tires can handle anyway (it's accelerating at over one g up to that point), and certainly it is all that is needed. With AWD and separate front and rear motors (which is the only sensible arrangement), shifting gears could be a challenge to manage well.
