Nothing prevents a dual-diaphragm figure-8 from having a highly symmetrical pattern, but it requires a willingness on the manufacturer's part to select capsules with very precisely matched front and back sensitivity, and--the part that evidently hurts too much for most of them--the willingness to reject capsules that don't have this characteristic. In the intensely price-driven market that most non-professional microphones (especially those from Asia) are part of, that's just not economically feasible. The manufacturers know that most people only use multi-pattern microphones as cardioids, so they don't risk (say) doubling the price to make the figure-8 pattern really good. However, for M/S to work, the side-facing microphone must be a true figure-8, with two lobes of equal sensitivity in opposite polarity to one another, and a response null midway between them, where little or no sound is picked up.
Some other microphone patterns, such as hypercardioid, are also "bidirectional" in that they have two lobes of opposite polarity with a "null" between them--but the lobes in those patterns don't have equal sensitivity, and the null occurs at angles other than 90 degrees. There's a direct geometric correspondence: The more sensitive the front lobe is relative to the rear lobe, the farther back the "null" occurs, and if the front lobe is less sensitive than the rear lobe, then the farther forward the null will be (plus then you basically have your microphone facing backwards). For a hypercardioid, where the rear lobe is 6 dB less sensitive than the front lobe, the null occurs at 110 degrees. For a supercardioid, with a rear lobe nearly 12 dB less sensitive than the front lobe, the null occurs at around 126 degrees. Theoretically you could imagine a cardioid as a "bidirectional" microphone with a rear lobe of 0 sensitivity (which is as low as it can possibly be), so the null is at 180 degrees (which is as far to the back as it can possibly be).
And if the two halves of a bidirectional capsule have unequal frequency response, then at each particular frequency the pattern will have a different shape, and the null will fall at a different angle. So the response at 90 degrees might be well nulled out at some frequencies but not others. That's bad for M/S as well.
Truth be told, though, nothing about a single-diaphragm figure-8 capsule design automatically guarantees it a symmetrical pattern, either; the single-diaphragm design removes some of the main variables that can throw things out of symmetry, but not all of them. Anyway, good single-diaphragm figure-8 condenser microphones are available from Schoeps, Neumann and Sennheiser for certain, and there may be others that I don't know about.
There are also reasonably small ribbon figure-8s from Beyer, Royer and more recently Audio-Technica that might be worth considering for indoor recording. They may offer better low-frequency response without electronic equalization--and low-frequency response is very important to the sense of spaciousness in an M/S stereo recording. On the other hand, a good figure-8 condenser microphone can be equalized so that its low-frequency response extends as far and as smoothly as you might ever want for music recording, since the response below the "turnover frequency" rolls off at a very predictable 6 dB/octave rate.
--best regards