There was a discussion on the EVDL about the losses in bearings. I commented that my Siemens LH5118 motor bearings seem to get quite hot, Victor replied to say this was normal and would get better as the grease got properly spread out.
So I ran the motor in at about 1000 rpm for a little under 2 hours:


We can see what might be a significant reduction in no-load losses after running in. It is odd that there is such a big spread, especially in the measurements taken after running in. I eliminated all points where the speed was changing by more than 20rpm/s -- it's quite difficult to control the speed since the throttle is a torque control and the rotational inertia is very small.
It turns out I made a mistake stripping data out of the graph where the motor was accelerating. The graph originally presented (click the thumbnail below) includes only data recorded while the motor was accelerating. The new graph you see above shows only the power at reasonably constant motor speed. After running in, you can see a significant decrease in the power requirements between 4000 and 6500rpm. Unfortunately I didn't collect enough data at higher speeds to draw strong conclusions but it appears that there was no significant change.
After the running in period the inverter reported 65C heatsink temperature and 25C motor temperature. When you run the motor at 10000rpm the inside of the splined hole increases quickly to from 20C to 50C.
The power recorded is the battery power, including the DC-DC powering the inverter and charging the auxiliary battery. I took 3.4Ah from the battery during the test, it's open circuit voltage dropped from 151V to 148V.
To do this test, I connected the motor rpm output on my Simovert inverter to the vehicle speed input on my EVision, I can do this directly because the EVision and the inverter share a ground. (Since I plan to use a fixed gear reduction I will be using this as my vehicle speed input). I set the various scaling factors so the EVision read RPM*100 instead of km/h. I pulled the speed, volts & amps out of the EVision telemetry using a program I wrote and plotted the results using gnuplot.
Full data available here