^ Yeah that.
A few things about it..
Practically: 90 degrees is easy to visualize and arrange when using an adjustable stereo mic mount.
Overall: When summed together, a pair of cardioids/supercards with a 90 degree angle between them produces a combined sensitivity pattern that's more-or-less the shape of a single forward facing subcardioid, which IME is generally optimal for recording in good acoustic situations regardless of the specific details of the stereo microphone configuration used to achieve it.
Stereo qualities: As Brad mentions, when we narrow the angle from 90deg and bring it in closer to PAS, we tighten up the combined sensitivity pattern, which increases the ratio of PA and stage sound (music) verses ambient/audience sound. When doing that, we can seek to retain good stereo qualities by increasing the spacing between the pair as the angle is made narrower (this is the basis for improved PAS). In doing that we increase the time-difference stereo qualities as a way of compensating for the reduction in level-difference stereo brought about by the narrower angle. The narrowed angle decreases level-differences between channels, while the increased spacing increases time-differences between them. Ok, nothing new there, but lets dig deeper into it..
When a 90-deg angled cardioid/supercardioid pair is setup with the appropriate spacing between mics, the stereo attributes of the resulting recording will consist of a pretty even balance of level-difference stereo and time-difference stereo. An 50/50 distribution of level vs time difference is a generally optimal target for a cardioid or supercardioid stereo pair and will be produced by a stereo configuration somewhere between DIN and NOS (a bit more toward the DIN side). However a 50/50 distribution isn't necessarily the appropriate target if using other polar patterns. To achieve the same near 50/50 distribution using a pair of subcardioids would require them to be setup at 170o/22cm to 180o/23cm. That will produce a combined sensitivity pattern that's basically omnidirectional rather than subcardioid shaped and might work great on-stage, but probably less so out in the audience even in a good room. If the subcards are angled less than that, we ideally need to use more spacing between them to compensate for the lack of level-difference due to the narrower angle, and the stereo qualities that result will be produced by time-difference more than level-difference. Omnis even more so. The stereo qualities that result from that are generally what we expect from those patterns. The opposite is true for a fig-8 pair. A coincident Blumlein pair with no spacing between the mics is 100% level-difference stereo with no time-difference and the stereo qualities that result from that are what are generally expected from fig-8s used as a Blumlein pair. To achieve a 50/50 distribution using fig-8's would require narrowing the angle to 70o with them spaced 34cm apart. That's not unreasonable for a pair of fig-8's in PAS. It should produce good stereo qualities and a collective sensitivity pattern that is tic-tac shaped, just as sensitive to sound arriving from behind as from the front, while reducing pickup of talkers and other sounds off to either side of the recording position more than any other configuration that produces roughly the same stereo image width.
A narrowed angle between a PAS pair increases pickup of the music and decreases pickup of audience which is the goal, but like Brad mentions, any obnoxious chatter off to either side which is picked up will tend to have less level difference between channels than if the angle was wider. If the narrowed angle is not compensated for by increased spacing, that chatter is going to increasingly compete perceptually with the music due to both inhabiting a similar position within the less differentiated stereo image. Increasing the spacing between the pair helps by perceptually pushing the talkers who were over to one side or the other farther out to the sides in the resulting stereo image keeping them a bit more out of the way of the music.
[Edit- I think the advantage of the increased spacing used by improved PAS is as much about getting the audience out of the way of the music as it is about improving the stereo image width of the musical content itself]
