Did the test again with a reference waveform (top track). I simply rendered the JS Tone Generator first, panned the track to one monitor only and played the tone while recording the same monitor's output.
TRS speaker cables had normal polarity (+tip/-ring/sleeve).
The monitor on the right is the one with the "Positive Phase" sticker.
100 Hz
6 kHz
So, as predicted, it appears the monitor that does not have the "Positive Phase" sticker is the one that has inverted phase. This test confirms (again) that both the tweeter and the woofer have inverted phase.
I think I'm going to just reverse the wiring to the drivers on the monitor that has the problem. There are clips that just pull off. Given they're a pair might as well make it permanent that they're in phase with each other.
The other test I did was a sweep, 5Hz - 20kHz, and I was going to look for irregularities in the waveform that might indicate the monitors are still not functioning according to spec (50Hz - 20kHz +/- 1.5dB) but I realized I don't have a microphone that has a flat enough response for that kind of test.