Oh, no, here we go again! Let me start out by saying that, after over a year of extensive experimentation, I have determined crank arm length does not matter very much. Though people range widely in size, almost all of them seem pretty happy with 170 mm crank arms. In the abstract it seems obvious that crank arm length should be proportional to leg length, but there doesn't actually seem to be much evidence for this.

That said,

Yeah, yeah, don't laugh! I haven't tried 135's; they're not readily available. But I have 127's on my Strida, 140's on the folder I rode today, 145's on my recumbent, 148's on my Downtube Mini, and so on. 127's feel very short. 140's are fine. I am pretty comfortable on all of them, but think best would be somewhere in the range 150-160 (see below). I too can tell the difference between 165 and 170, but anything over 170 feels pretty much the same to me: too long!

For reference, I'm 6' tall.

By the way, I have read the article, but am not sure I followed Sheldon's reasoning, so I am not going to pretend I am representing his view here.

If you take pedal speed into account when calculating gearing, then crank arm length makes a big difference. A longer crank arm means a larger pedaling circle, so at a given gear ratio and a given speed, the rider's foot must be going faster if the crank arm is longer. With me so far?

But if that is obvious (and I think it is), it is not necessarily obvious that pedal speed has any relevance to anything. The muscles pushing the pedal are in the thigh, not the foot, so pedal speed may be completely irrelevant. With the longer crank arm, the thigh has to move a little farther, so a little faster; but whatever movement it makes, it makes that movement at the same frequency regardless of crank arm length.

It is also obvious that longer crank arms give better leverage; but this is only relevant at low cadences. If you have a wide range of gears to chose from, you can maintain a high cadence, so you don't need that extra leverage at all.

Short crank arms make it possible to spin at higher cadences, while long crank arms make it possible to mash at low cadences. While it is possible to mash a short crank arm, it is painful (and presumably bad for your knees); and while it is possible to spin long crank arms, this is initially tiring and painful in the long term.

My impression is that the chief disadvantage of a longer crank arm is the unnecessarily wide range of movement of the thigh and both hip and knee joints (which may in turn be bad for those joints). When I ride long crank arms, my butt gets very sore. You may argue that this is only because I'm not accustomed to the longer arms; and you may be right. I don't know.

My current thinking is that crank arm length is an easy way to find a balance between aerobic and anaerobic effort. If, after a long ride (100 miles, for example) your legs feel like jelly but your chest feels fine, then your crank arms are too long and your cadence to low. But if, at that point, your legs feel fine and your chest hurts, then your crank arms are too short and your cadence is too high.

A couple Saturdays ago I rode 140 miles on my touring bike, with 152 mm crank arms. I was really tired by the end of that ride, but I must admit my chest felt worse than my legs. So maybe I'd be better off with something a little longer....

No it's not!Not exactly! It will definitely reduce pedal strike and toe strike. It will not increase the gear, but it will enable you to go a little faster. On my Strida, where the speed tops out somewhere around 16 mph and my feet are spinning insanely fast, my super short crank arms (127) help; I can spin them a little faster, and I can therefore go a little faster. But while I do get a rather neglibible increase in top speed, it comes at a significant decrease in accelleration, especially from a stop. On the Strida it has the additional benefit of increasing knee clearance; for a person of my size there's a constant danger of hitting the knee on the handlebar, which can cause erratic steering!