I'm baffled that people still don't seem to understand why it might be a bad idea to run the smallest cog with the middle ring.

Sheldon does not say this because of the fact that the chain isn't straight. He doesn't say it because of any other principle that applies to derailleur bikes only.

imagine you had a derailleur-equipped bike that was designed to use only these two chainring-cog combinations:


56/14
36/9

All three gear ratios are the same. (4 revolutions of the wheel per rev of the cranks)
The 36/9 puts the largest amount of tension on the drive side (top) of the chain. The reason is that the part of the chainring pulling on the chain acts as a lever with a length (from the bottom bracket center) of roughly 3 inches. Your foot is pushing on a crankarm that's 6 inches or so, and if you put 150 pounds of force on that crankarm, you're putting about 300 pounds of force on the chain.

Contrast that with a 56-tooth chainring. (radius of 4.5 inches or something.) Put 150 pounds of force on the pedal, and you end up with about 225 pounds of force on the chain.

It's a matter of leverage. the ground "pushing" back on your wheel can transmit more force to the chain through a small cog. the crank can transfer more torque to the chain through a small chainring than through a large chainring. add up the competing forces and you get more chain wear.