I was reading this paragraph and it got me thinking:
The closed ends of the honeycomb cells are also an example of geometric efficiency, albeit three-dimensional and little-noticed. The ends are trihedral (i.e., composed of three planes) sections of rhombic dodecahedra, with the dihedral angles of all adjacent surfaces measuring $120^o$, the angle that minimizes surface area for a given volume. (The angle formed by the edges at the pyramidal apex is approximately $109^\circ 28^\prime 16^{\prime\prime}$ $\left(= 180^\circ - \cos^{-1}\left(\frac13\right)\right)$
This is hardly intuitive; is there a proof of this somewhere?
If you want to divide space up into uniform volume cells with minimum surface area, the honeycomb is not optimal. Look at the Weaire–Phelan structure. While honeycombs are not quite optimal, they are certainly close enough for bees -- they're suboptimal by only 0.3%.