Am I right in assuming this box flipped polarity on both channels at a time?
I have noticed a number of times that phenomenon. It was always traced down to the fact that loudspeakers have asymmetrical response. It is particularly sensible on bass units, where the diaphragm excursion is important. The main reason for that is that the induction varies differently in a direction than in the other.
Most loudspeakers are subject to that. Loudspeaker designers have tried since ages to reduce this phenomenon. This is the base for JBL's SFG (Symmetrical Field Geometry) or Purifi's Constant Force Factor.
There are other factors that result in asymmetrical response, surround radiation, air compression vs. deflation.
What is maybe missed by the average listener in evaluating speaker performance is what goes on when sound is recorded and how “attack” can be just as much from incursion as excursion from a speaker diaphragm. For example with a drum kit - the kick drum mic is put into a hole in the front skin (the side facing the audience not the drummer) behind and pointing at the back of the back skin where the beater strikes and the snare and toms and percussion mics are place above the skin surface, opposite to a kick. So the kick being initially positive pressure on strike it would appear that the other skinned instruments would be at negative pressure on strike - the fundamental mode 0,1 and the 0,2 mode contribute to the “thump” sound of a drumskin - the 0,2 at a point 0.436 of the distance to the edge from the centre has the opposite direction of travel, due to the nature of sound wave propagation on a circular membrane the subsequent contra rotating interacting waves at coincidence cause sequentially, compression and rarefaction of the air at the other vibrational modes of the skin and in different patterns according to the mode.
The mic in the kick drum is at a point where the initial impact of the beater provides a distinct positive “push”, for the other top miked drums the position of the mic can determine the polarity - aim it directly at and over the strike point in the centre you’ll get negative pressure, move out nearly 1/2 way to the side and further where a mic will normally reside and this reverses to the initial skin travel direction mode 0,1 negative pressure combined with resultant mode 0,2 positive pressure, which has a longer decay time but radiates less, plus of course the subsequent wave propagation around the skin - as the mic would get in the way of the drumstick it has to be placed towards the outer rim and so is also more subject to the mode 0,2 positive pressure on strike notwithstanding the initial fundamental negative pressure impulse. Looking at a top miked drum first impulse waveform there is usually an initial negative going travel followed immediately by a larger positive going waveform.
The tone of the drum comes from other modes of skin motion which contribute to the pitch and decay time of the vibration of the skin, modes 1,1 2,1, 3,1 4,1 and 5,1 in ever increasing complexity of reversals of skin direction.
Yamaha produces drivers using ceramic coating to rigidify speaker cones to negate the effect of ripple distortion and cone resonance.
Considering that no matter which mic you’re listening to that immediately after a positive pressure impulse there is a corresponding negative one (or vise versa) then positive and so on then it becomes apparent that not only is it important that the mic responds in kind but also the speakers and its enclosure must do exactly the same.