In the many measurements I've done, a 180 degree polarity reversal will cause a suckout of about 6db at the crossover frequency in cases where the woofers of the mains and subs are placed next to each other, when observed on a RTA. The farther apart the placement, the less the suckout will be, but depending on the place where you are measuring, there will be random peaks and nulls because of the acoustics of the room. These peaks and nulls will occur whether the woofers are in polarity or not; they will change in frequency somewhat from in polarity to out of polarity.

While all this is obvious on the screen of an RTA, the audibility with actual program material is generally much less. In addition, changing the polarity of the sub will yield flattest response in some locations and worse response in others.

Generally, I would determine the setting that yields flattest response at the listening position and call it good. Trying to find a setting that is valid in all parts of the room is difficult, and it will get harder the farther apart the sub is placed from the other speakers.