It sounds like it could work, but I suspect that unless there was a probe inserted into the exhaust stream there would be little usable information coming from an exhaust header surface temperature reading.
First off, I would think that the temperature changes would be very small over the range of idle mix settings I typically encounter when balancing the throttle bodies on my bikes.
Second, there is probably some thermal transfer lag as the heat travels through the walls of the headers. This would require a lot of wait time while the tuner waits for the temperature to stabilize.
Last, I would expect the temperature readings on the headers to be easily disturbed by the slightest of air currents. then there is the problem of finding the spot on the header where a usable temperature reading could be taken.
Personally, if I was to use header temperature as a tuning parameter, it would be at higher load/rpm regions of the operating range.