Even simpler to explain how full frame could work, with unclipped corners, using the M mount: the imaging light is projected from the lens to the sensor from the last element at the back of the lens, which is, I think, always situated slightly behind the plane of the lens mount, not at or in front of the plane of the lens mount. Simple. All other explanations are surely true, but none, other than the above, is necessary.
There are problems with designing lenses this way, though. The closer to the sensor plane the backmost elements are (or, for that matter, the exit pupil is), the more angled the light has to be to reach the periphery of the sensor. The larger the sensor, greater that angle is. Highly angled light doesn't flow into the pixel wells (for FSI designs), and it is difficult to create microlenses with a great enough power to bend the light back into the well. For BSI designs, the high angle of light results in a significantly greater amount of reflection rather than refraction, so the light is simply lost. This increases vignetting in the corners. The Sony FF mirrorless options have this problem. Sony has tried to mitigate the issue by using differently designed microlenses in the periphery, however it is only a mitigation, not a solution to the problem.
Another problem with lens elements being mounted so close to the sensor plane is ghosting. A lot of ghosted light that reflects off the sensor is so dim that the inverse square falloff law results in it being invisible, for all intents and purposes, once it reflects off the back lens elements and back onto the sensor. With a much shorter sensor to back element distance, ghosting becomes a much greater problem. This actually occurs with most mirrorless designs today, including the EOS-M and Sony A7 series.
As much as everyone seems to want smaller and smaller and more compact cameras, making cameras that way has it's tradeoffs, it's cons. There isn't anything simple about creating pancake lenses that could work ideally for full-frame sensors in a mirrorless design. The benefits of using larger camera bodies with larger flange-to-sensor distances is you don't have these problems. A large flange distance, such as 44mm for Canon DSLRs, means light, even for a FF sensor, never has to reach significant angles, making microlensing on the sensor far more effective at guiding light down to the photodiodes. The greater distance results in a longer distance for reflected light to fall off and not cause ghosting.
I think it will be interesting to see how a FF Canon Mirrorless fares with purpose-built lenses. I suspect we'll see many of the same problems that the Sony FF mirrorless cameras experience. Canon has superior lens design capability vs. Sony, so in the long run I think they could build better lenses for a FF mirrorless system...but there are physical limitations for lenses just as much as there are physical limitations for sensors.