Most headphones, despite the back emf, resonances, quirks in mechanical and acoustic properties (e.g. electrical impedance can measure differently when clamped on head vs. not), the fact that you're dealing with a coil, etc. actually still are pretty resistive throughout the audio range. At higher frequencies than audio, I guess it looks more inductive than resistive, but not as much under 20 kHz. You can see the impedance sloping up at high frequencies at some headphones.
I'm not sure what you mean by this:
[quote]When you see FR graphs of devices going all crazy with resistive loads, that can't be accurate.[/quote]
Maybe some amps are marginally stable and misbehave with some non-resistors, like maybe headphones connected with a very long (capacitive) cable.
The biggest thing to look out for is the output impedance, which can cause easily non-trivial differences in frequency response. On some sets it results in a little higher nonlinear distortion as well.
But anyway, the reasons for using resistors are outlined in the first two replies. Anyhow, it's been suggested that some amps measure great into resistors but don't look as good with actual headphones, but when pressed to demonstrate that (hook up headphones as load instead of resistors and measure that), people don't seem to be able to demonstrate that happening.