I assume that given the variety of program material there is no hard and fast rule...
Now I'm not an expert on power amp design, but I'm a student of tripping circuit breakers. I'm sure others here will pitch in with more details, but here's my spin on this.
I would say that your two main factors to consider are #1) amplifier power supply technology and #2) music crest factor.
#1) Old school amplifiers used full-wave bridge rectifiers on a big transformer with big storage capacitors. These full-wave bridge rectifiers only drew power from the 60-Hz power line on the peaks of the sine wave. And those peaks could trip a circuit breaker pretty easily. Plus the output transistors were often biased in class AB2 mode (if memory serves), which causes a lot of output transistor heating as a trade-off to better linearity. However, modern power amps using Class-D circuity draw power during more of the 60-Hz sine wave, and thus can avoid the peak amperage draw of old-school linear amplifiers. Plus transistors in Class-D (On/Off) mode have less waste heat produced, drawing less power for the same program material.
#2) Different music will have wildly different crest factors (ratio of the peak level divided by the average level). So if you're doing reggae or heavy metal, the crest factor can be quite low, and the average level of the music will be very close to the peak level your amplifiers can produce. That's when you'll be tripping circuit breakers. This crest factor is also reduced by processing such as dynamic compression, which ups the average of vocals and bass guitars. Other music, such as jazz, will have a very high crest factor with lots of transients. In that case you could run many thousands of watts worth of amplifiers from a single 20-amp service. The peak power is there to give that style of music its dynamics, but doesn't need much average power from the wall outlet. In any event, a power amplifier reproducing music will rarely be subjected to producing full RMS power, except for perhaps some club tunes. Your mileage will vary, as they say in the auto biz.
I did a few experiments with amplifiers and clamp-meters in my youth, but not enough to really learn anything useful. However, with today's modern clamp meters from Fluke and Amprobe (and others) it would certainly be interesting to monitor the RMS amperage draw from your amp rack. Just remember that you can't simply clamp an ammeter around the power cord. You need a special cable with the hot and neutral separated so that the line amperage doesn't cancel itself out. I don't have time to do such a study right now, but it sure would be interesting.