Huh???
What makes you say that "all of the amps limit"?
Speakers are not purely resistive loads. The average impedance is far higher than the nominal and even so 3000 watts or so of true average power into a speaker is a TON of power. You can clearly see trends in the sub compression during the sweeps well before the loudest sweep.
There are only a few cases where the amplifier limits during the sweep. Most of them are extremely high power multiple driver cabs (MAUL, Skhorn) or systems with an impedance too low for the bridged K20 ( 1 ohm or 2ohm nominal load,( RF-19, dual 21IPal, SP4 18D1)) or where the SP1-6000 has not enough power. Even in those cases ONLY the loudest sweep runs into amp limitations. In those few cases what is the alternative amplifier that will provide an increase in headroom over a bridged K20 or a SP1-6000? In what case would MORE amplifier than this be put on a single cabinet?
The speakers are tested with both a very long duration signal and also very short duration burst signals. The data is there to make judgments on how the cabs respond to both and what will happen with normal signals that are in between.
The long term output sweeps do not show heating of the voice coil only. They also show the effects of suspension changes and loss of excursion linearity beyond Xmax as well. There may also be some amplifier limiting in the loudest sweep for passive cabs. All of these things are shown in those graphs.
If you want to see the effect of thermal effects only look at the repeat lower SPL sweep taken after the build up to the maximum sweep level. This is the one that shows how much shifting and heating has occurred in the driver with the amplifier and excursion related changes removed.