You can perform a stability test like this, but you have to be careful where you inject the test signal. Basically you want it to be at at a node which sees a very high impedance in one direction. That document shows doing the injection between Vout and the feedback attenuator, which can work if the attenuator impedance is high. Another common place to do it is between the error amp output and the modulator (usually the signal is a lot "cleaner" there than the output), or even in the feedback network of the error amp. But there's no reason it doesn't work at high power levels.
As for the resistor, it's not really critical, it just has to be much less than the normal impedance of the insertion path, so that it doesn't affect the transfer function by being there. So if your attenuator impedance was 10K, then you would probably want much less than 100 ohms. But if Rs is too low your injected perturbation will be very small.