In days of expensive tubes, adding some extra open loop gain would have been expensive, thus only mild negative feedback was used. The tubes are reasonable linear, having similar distortion properties and microphones and speakers in those days.
The whole reproduction chin had a nice S-shaped transfer function, reducing the gain gradually, when the signal amplitude was increased, thus acting as a gently compressor. Thus, short amplitude peaks due to adding two or more tones, was gradually attenuated, but general appearance was maintained.
With not so linear transistors, but with a lots of cheap open loop gain available, the obvious solution was to use a lot of negative feedback to have a stable closed loop gain and thus a large linear region. Everything works fine, until the available voltage range (typically the rail to rail voltage) is reached, after which, any peaks are hard clipped, causing nasty high order harmonics.
When two or more pure tones are added, short high amplitude peaks are produced, even when the average output voltage remains the same. Thus the peak to average increases with added tones.
In order to avoid nasty clipping distortion, the transistor amplifier must have so large peak to peak output voltage range that it never clips.
With a tube amplifier, the short and high peaks can be gently compressed, without causing too much discomfort (and some even prefer signals distorted in this way), thus the input voltage can be increased, increasing the average power output.
To exaggerate a bit, a transistor amplifier should be dimensioned according to the peak power, while a tube amplifier can be dimensioned closer to the average power. Thus, a 30 W tube amplifier can sound as well as a 100-200 W transistor amplifier.