I'll add some personal insight I've gained over the last year and a half or so from using an ambisonic mic with user adjustable polar patterns. For those unfamiliar, the system allows the user to adjust microphone patterns and angles after the recording has been made, while listening to playback.
Keep in mind a few constraints-
I've used this primarily for on-stage recordings, and I think perhaps more importantly,
the system is restricted to coincident mic configurations only.
Given that, I most often find myself settling on a polar pattern that lies some where around hypercardioid as defined by John's post above, with an angle between mics of anywhere from 110 to 120 degrees. Generally, and as you'd expect, as the polar pattern is adjusted more towards figure-8, I narrow the angle between microphones to something closer to 90 degrees (classic Blumlien) or even less. Likewise, as the polar pattern is adjusted to be closer to cardioid, a wider angle becomes preferable. My primary observation about that aspect is this: For coincident arrangements, the optimal combinations of pattern and angle are usually those in which the primary axis of one microphone is approximately aligned with the angle of maximum rejection of the other.
Given that, the decision on which of those particular combinations of pattern & angle is best is usually primarily determined by optimizing the direct/reverberant balance and degree of 'openness' in the spatial timbre of the recording (cardioids sounding drier, more direct and closed at one extreme, excluding the subcarioid to omni end of the polar spectrum, and Blumlein figure-8's sounding more reverberant, 'airy' and 'spatially open' at the other). Secondarily the choice is determined by the stereo spread and imaging of the direct sound sources, but I find those aspects can be regained at other combinations of pattern and angle, whereas the reverberant balance and spatial openness cannot.
Again I emphasize that this applies specifically to coincident mic setups. Near-spaced and wider arrangements will differ by trading angle between mics for more distance between them. To a lesser extent, these observations may be specific to the scenarios in which I've made these recordings, yet I feel comfortable extending the observation to coincident mic arrangements in general.