Typically a full bandwidth signal (pink noise for a logarithmic scale chart and white noise for a linear scale chart) is used. No serious review site uses sweeps. This would take ridiculously long to do (since to get a high resolution plot you need many, many points). These are undoubtedly more complex than music, I would assume, though I have no proof other than it simply having output at more frequencies at the same time than music. I see and respect the logic behind your arguement, but because pink noise is used instead of single frequencies, I believe it does not apply in this situation. Also please be aware that since all frequencies are used, there will be differences in phase as a function of time due to the different frequencies having constructive and destructive interference with eachother. That would prevent the amplifier from having an "easier" job with a full bandwidth signal compared to a test tone or music.
