The theoretical stuff-
Consider an idealized case. Imagine just the left two mics to simplify things; one cardioid and one omni. If the mics were coincident and the frequency responses of the mics were the same and the patterns were perfect and the levels were just right and the polarity was reversed on one of the mics, and you then summed the signals you would have the equivalent of a single figure-8 pattern.
Why? Mathematically, a cardioid pattern is the sum of an omni and a figure-8. Adding a polarity reversed omni and cardioid signal together would cancel the two omni components since one of them is the opposite of the other leaving you with a figure-8 pattern. -(omni) + (omni + figure-8) = figure-8
If the polarity was not reversed on one of the signals then you would end up with a wide cardioid. (omni) + (omni + figure-8) = 2(omni) + figure-8 = wide cardioid.
Any spacing between the mics introduces phase shift as someone else mentioned previously, and that makes things more complicated. The amount of phase shift is not constant but increases with frequency. When the phase difference between signals equals 180 degrees or any multiple of that, the signals will destructively interfere. If the signals are exactly the same amplitude at that frequency, then they will cancel completely. Conversely, when the phase difference between signals is zero or 360 degress or any multiples of that, they will constructively interfere and increase in level by 6db. Phase differences between those extremes will interact too, according to their levels and phase differences.
All that means that you aren't just changing the polar pattern, but changing the polar pattern and the frequency response and the phase and impulse response in various ways at different frequencies.
The real world-
With the two pairs of mics spaced a small distance apart, frequency cancellation due to inverted polarity will mostly likely be audible as a change in level in the lower frequencies (where the phase shift between signals introduced by the spacing is small). You'll hear changes at higher frequencies too because of the complex comb filtering intereaction, but the combing is dense and complex up there so you won't hear it as a reduction in level. Instead you may hear a change in timbre, width or depth, or in the transients. Regardless of which polarity you choose the comb filtering is going to happen up there anyway. Understanding that, either polarity might 'sound better'. For example- destructive interference at low frequencies may sound good by acting as sort of a high-pass and reducing excessive bass, likewise comb filtering higher up can add a desirable shine and air to the mid and high frequencies.. or not. What sounds good is good.
I'm guessing the 1st link in the original post is with the mics in the same polarity (in other words, one file had the polarity flipped in the DAW) and the 2nd link is the unadjusted files with one pair out of polarity with the other. I prefer the 1st becasue of the richer quality of the bass, because I hear depth in the recording more clearly and because the higher transients seem less smeared and less forward. That's with earbuds in my ears from the computer sound card. Of course YMMV, I just wanted to clarify what's going on. Hope that helps, if not ignore all the technical BS and just do whatever sounds good.