Gutbucket, in stereo shotgun microphones, the capsules for M and S (in every example I've seen) are at the same point within the housing behind the interference tube. Instead of a unitary figure-8 capsule, a matched pair of side-facing (outward-facing) cardioid capsules is used, whose outputs are summed "in antiphase" to form a figure-8 pattern. These capsules are positioned on either side of the central, forward-facing capsule that joins to the interference tube.
In the attached "exploded view" of a Neumann RSM 190 or 191 (same mike, different matrix box), the capsule assembly is the cube-like piece in the middle (resistance is futile, so maybe let's try capacitance instead?) labeled with position numbers 5 and 2.
The whole way that M/S works REQUIRES the path length from the center of the direct sound sources to the "M" and "S" microphones to be the same, since the reinforcement vs. cancellation between the M and S channels is what produces the L and R signals. The effect of any discrepancy in distance would be inversely proportional to the sound wavelengths. At low frequencies there is room for some error. But stereo localization depends greatly on upper midrange content (say, from 2 to 5 kHz), where an inch is a significant fraction of a wavelength.
--best regards
P.S.: Nick's Picks, you only think you don't care whether the "M" capsule is single- or dual-diaphragm. Dual-diaphragm is fine for the patterns "beyond cardioid" (supercardioid, hypercardioid, figure-8), but a dual-diaphragm cardioid is only a cardioid in the midrange (and if it's small, perhaps at higher frequencies as well, depending on some other factors). At low frequencies the pattern invariably "blooms out" toward wide cardioid. That's OK in normal studio use, where the added room pickup at low frequencies (say, on a vocal soloist) can increase the sense of roundness and warmth. (Probably the reason large-diaphragm condensers have a reputation for "big" sound in the low frequencies is that nearly all large-diaphragm condenser mikes use dual-diaphragm capsules.)
But in a stereo microphone, whether X/Y or M/S, that characteristic is the opposite of what you want, because it causes the eventual L/R output to "blend toward mono" at low frequencies. That was useful in the era of stereo LPs, since difference information translates into vertical groove modulation, which had to be carefully controlled. But it's not at all desirable in the era of digital meda. The usual main problem with coincident stereo recordings is a lack of spaciousness, and a widening cardioid(-ish) pattern at low frequencies goes directly to the heart of the problem and makes it significantly worse. A good single-diaphragm cardioid, on the other hand, will still be a real cardioid at 50 Hz and even lower.
Attached is a set of polar diagrams for a classic dual-diaphragm microphone (Neumann U 47), which had pattern settings for "omni" and "cardioid". The ones for the "cardioid" setting are on the left side of the diagram; notice how the pattern spreads out more, the farther down you go in frequency. Not all dual-diaphragm microphones have it this bad, but this type of capsule has many, many imitators among manufacturers who aim at the studio market.