nedstruzz, this has partly to do with membrane tension--the greater the tension on the membrane, the less sensitive it will be to wind, breath noise, popping from spoken or sung consonants, and mechanical shock. Wide cardioids such as the Schoeps MK 21 are mostly pressure transducers; their characteristics are rather close to those of single-diaphragm omnidirectional microphones. In general (and this is quite a generalization), all other factors being equal among single-diaphragm capsules, omnidirectional capsules tend to have the highest membrane tension--while at the opposite end of the spectrum, figure-8s tend to have the lowest. So that's one major factor.
Another factor comes from the fundamental way a directional microphone works--it responds to the moment-to-moment difference in sound pressure in the air at two points which are adjacent, but separated by the diaphragm of the microphone. The paths that the sound waves have to travel to reach those points are separate and (in all cases but a figure-8) unequal. Anyway, sound waves reach both sides of the diaphragm in directional microphones, with the timing (phase) relationship determining whether they will reinforce or cancel each other when they get there. The timing relationship depends on the angle from which the sound is arriving, which is how the microphone can emphasize the pickup of sound from one direction while (relatively) suppressing the sound coming from another.
OK, that's how it works when the energy reaching the microphone comes from sound waves--but what about wind? Wind can be steady and unidirectional, or it can be turbulent, or it can be anything in between; let's not go into those details, thought they certainly matter when you're choosing (or designing) a windscreen. The point is that any pressure-gradient microphone is effectively "comparing" two energy sources as they reach the front and back of the diaphragm, and with sound you are always comparing a more or less phase-shifted version of the same signal with itself. But with wind, the two "signals" being "compared" often have nothing much to do with one another. The difference between them is more like random noise--so noise is what you get at the output.
Meanwhile with a pressure (omnidirectional) transducer, the front and back of the membrane aren't both exposed to the air--only (usually) the front is, and the capsule is sealed. With wind you may well have random-ish "signals" hitting the front of the membrane, but you don't also have uncorrelated, random-ish "signals" hitting the back of the membrane at the same time. So it's like the difference between one person who has no sense of rhythm whatsoever banging on a can, versus two people who have no sense of rhythm whatsoever (AND who are paying little if any attention to each other) banging on two cans.
--best regards