The negative polarity of the midrange unit ensures that when the phase response is also considered, the outputs of the drive-units sum in phase in the crossover region to give a flat amplitide response on the intended axis.
...
Because then the drive-units would not sum in-phase and you would get a large and very audible suckout in the crossover region. Pathologically poor design.
So the drive units sum up in phase in the crossover region - fine; but isn't then the midrange driver always out of phase within the rest of its operating range, ie. within those crossover points? And isn't that a bad thing? The response above seems to imply that a midrange (in a 3-way design) must always be wired in reverse polarity and that no one can design a crossover with them in-phase and avoid suckouts?
It feels like I am missing something here...