Some more in-depth dirt on phase interactions between channels-
Fundamentally, there are two basic strategies for reducing potentially problematic phase interactions between microphone channels, and they are essentially opposite approaches to each other. One seeks to eliminate the phase differences between channels by placing both microphones in a single "coincident" point in space, making the phase relationship between those channels identical. The other seeks to make the phase relationship between channels different enough that they will not interact problematically when summed, achieved by spacing the microphones far enough apart from each other that the phase relationship between them is essentially randomized. That's the basis of the 3-to-1 spacing rule for close microphones that will be summed together into a single channel (Note- the 3-to-1 rule is not intended to apply to stereo microphone configurations where the channels are not summed. Many sources erroneously state that it does apply to stereo mic'ing, but a stereo pair is not summed, and 3-to-1 is geometrically impossible to apply anyway). A complication of the spacing approach is that the phase relationship between spaced positions varies in a complex way. It varies with spacing distance, by frequency, and with the angle of arrival of the wavefront, and all three aspects interact with each other. For wavefronts that arrive from any angle other than very close to perpendicular / directly on-center, the phase relationships at the high end of the frequency range is essentially randomized down to an increasingly lower frequency as spacing is increased. The greater the off-axis angle from which the wavefront arrives, the greater that difference becomes. The difference varies with frequency as well - there is increasingly less phase shift as the frequency decreases.
When two similar signals with different phase relationships are summed, at each multiple of one full frequency cycle or 360 degree of phase difference the signals will constructively interfere to the greatest degree, and at each half cycle or 180 degree difference the signals will destructively interfere and cancel to the greatest degree. Below some frequency where the difference in path length to each microphone is close enough to the wavelength of the frequency in question, the phase difference between channels will be less than 360 degrees and no longer effectively random in a perceptual sense. At low enough frequencies with significantly long wavelengths in comparison to the spacing between microphones, the phase difference will be less than a 1/4 wavelength and the two signals will essentially remain in phase, acting effectively like coincident placement below that frequency. That may or may not help explain how spacing a pair of omnis more widely can create more low frequency difference between channels.
How does all this relate to why a pair of wide spaced omnis with a coincident pair in the middle might be a highly effective four microphone arrangement for taping from an audience position? And how might it be applied to deciding how to arrange it? Consider the different qualities of direct-arriving verses reverberant sound. Direct-arrival sound requires a more highly correlated, or identical phase relationship between channels to present a well defined stereo image that features directional clarity and well-defined imaging, while reverberant sound has a more highly-randomized phase relationship and requires the preservation of that between channels to sound open, airy and natural rather than closed-in, flat and monophonic. A recording will translate better if we can arrange things so as to preserve or even enhance both of these relationships, however the two are at odds with each other. When using a near-spaced pair we are finding something of a best middle-way compromise solution that may not be optimal for either but works well enough for both.
When using a coincident pair in combination with a wide-spaced pair we can set things up so as to have each pair better optimized for what it can best contribute. A coincident pair produces no phase difference between channels and does a great job in translating clear and distinct directional imaging of direct arrival sound. A wide omni pair randomizes phase across a significant portion of the frequency range, which does a great job of translating a big lush open reverberant sense of space, and at lower frequencies can provide useful non-random phase differences that makes for more enveloping, non-monophonic bass. There is still a compromise to be made however, although now it is a different one - managing the phase interaction complexity between both pairs. Using a coincident pair in the middle helps with that, as does spacing the omnis far enough apart that they won't produce objectionable phase interactions with the coincident center pair (yet not too far), and a somewhat wider spacing that can be used when a center microphone or pair is included also helps with portrayal of the reververant aspects of the recording.