Or, maybe, it reflects the fact that this format simply works. It produces balanced, pleasing images of a large variety of subjects.
I agree with N. Devlin (and Richard Sexton?): the 3:2 shape has never been the dominant or "middle of the road" choice for prints larger than snapshots, or for paintings or drawings in any era. Instead there has been for many centuries, in painting, drawing and photography, a dominant range of shape choices ranging from about 5:4 (1.25) to 3:2 (1.5), with the most common choices being in between, about 4:3 (1.33) to 7:5 (1.4). The 3:2 shape is at one end of this main range, not in the middle: shapes wider than 3:2 have always been far less common than shapes a bit less wide like 7:5 and 4:3. (And please, no mystical nonsense about the Golden Ratio, which has never been at all a common choice for "rectangular art works".)
The 3:2 shape of 36x24mm made sense with the constraint of 24mm width of the dominant 135 format film, but has never had much popularity in any other film format or for paintings or drawings. for example, where are the 3:2 formats in MF or LF cameras? There was no major obstacle to making either a "9x6cm" or "6x4cm" camera using 120 or 220 roll film, but the first was rare and the second never happened AFAIK.
The 3:2 shape is rather clearly film-era historical baggage from the facts that:
- 135 format film, 24mm wide, became entrenched with huge cost advantages over any competing format for rolls of film.
- With the short edge constrained to 24mm, the most flexible option was to make the frames rather long in the other dimension, to cover all the most popular shapes with only horizontal crops needed so that you got a good range of shapes while still using the full 24mm frame height. With the dominant shape range from 5:4 to 3:2 it made sense for 135 film cameras to use the largest and widest of these: 36x24 for 3:2, rather than say 32x24mm (4:3) or 30x24mm (5:4).
But with sensors, that 24mm constraint is gone(*) and instead the most efficient use of resources is to choose a shape near the middle of the range of most common shape choices, to minimize the average fraction of the image lost to cropping when wanting a different shape. That is, something strictly between 5:4 and 3:2, like either 4:3 or 7:5. And of those, 4:3 has clearly won.
By the way, 7:5 might be a nice compromise, because it is an almost perfect match to the ISO standard "A" paper sizes, and still closer than 3:2 to 10x8, 14x11, 20x16, etc.
(*) The 24mm constraint is not entirely gone, because there are still DSLRs that use lenses and lens mounts originally designed for the 36x24mm film format, and these still have the 24mm height constraint (from VF mirrors and baffles within some lenses for example), so 36x24mm is still entrenched in "full frame DSLRs" for historical reasons.