What is not obvious is that pedalling forces from the right leg don't go through the shaft. Your foot presses on the crank and the crank is connected to the spider and that to the chain. No force goes through the shaft when your right leg pushes down. When your left leg pushes down, the force goes into the left crank, through the shaft, into the spider etc etc. The stiffness (resistance to twist and resistance to bend) of that shaft is determined by the shaft's diameter.
So what if it bends and twists a bit, I hear you say? Well, you don't care but the bearing cares. Firstly, it pulls the bearing's inner race out of alignment with the outer race. If you imagine the balls running between two tracks which are perfectly aligned, and you twist the tracks so they are not running at angles to each other, you can imagine what that does to the bearing balls - they fail prematurely. They fail at a ratio 3 to 1, left to right. Compared to similar bearings used in industry, they fail in a much shorter service life.
Twisting forces have an effect on the bearing/shaft interface too. Shimano anticipated that and inserted a plastic sleeve between the two. That way any twisting and bending is taken up by plastic deformation in the plastic sleeve. It also eliminates noises. SRAM's left hand side is metal on metal.
Again, the movement between bearing and shaft may seem trivial, but just listen to BB30 owners and you'll hear howls of disappointment.
Look at this thread here:
https://www.cyclechat.net/threads/cannondale-bb30-bottom-bracket.196684/post-4207483
@Globalti posted a nice picture there of his BB30 crank suffering from bearing fretting. Compare right to left and note the difference. Cannondale, the lead inventor of BB30 thought that they could get away with an aluminium shaft if they increased diameter to 30mm. Wrong!
Again, the forces between the bearing and shaft are not obvious. They're simultaneous bend and twist forces. The twist forces cause shaft/bearing contact points to shift not radially, but axially.
