This is the most encouraging part. I had the propellers printed with two materials, Shapeways' "White, Strong, and Flexible (polished)", and their "White Detail". Both were able to spin (at about 8500 rpm) without any issues, and both were very well balanced -- only one needed a slight bend to track level. The vibration seemed lower than with my other props, though the rpm was lower as well.
I tested hitting the spinning blades against a wooden beam. The WSF blades were impossible to break like this! (They seem slightly more flexible, and a bit gummier, than the molded blades, which helps them here.) I didn't even notice leading edge damage. The "White Detail" was a lot more brittle, though... this would happen on almost every impact:
My model had a 0.9 mm diameter hole, to fit snugly over the motor's 1 mm shaft. In WSF, the hole fused shut, so I had to drill it out with a 0.7 mm diamond bit. (The hole must have guided the bit, though, because it was still balanced despite my non-precision approach to drilling.) In "White Detail", however, the hole was too large, and one of the four props would fly off under even its modest 5 grams of thrust.
This was disappointing. My quadcopter weighs 60 grams, but all four of these props only provided 20 or 30 grams of thrust. (For comparison, the pre-made purple props, with too small a diameter, were able to lift it, barely.) They also only got up to 8500 rpm, as opposed to 13500 rpm for the purple ones.
The first problem is that the blades are stalling. I had just eyeballed the angle of attack, which blended linearly from 30° at 25% radius to 10° at the tip, for a geometric advance ratio of 72% However, after doing the math on the induced velocity I need to hover, I really need about 29% no-slip advance ratio, which ends up being thirty-something geometric depending on angle of attack.
Next problem is aspect ratio. There's a bit of a tradeoff between chord length and angle of attack, but I've heard that at low Reynolds numbers (I'm around 20,000), it's better to have longer chords flying at lower angles of attack and lower lift coefficients. Most of the commercial props have an aspect ratio between 3 and 4, but mine was 6, so I'll lower that next time.
Finally, airfoil shape. Before, I just had a slab with a thick leading edge:
I'm missing a lot of data: I know how much thrust I need, but I don't know the efficiency vs. rpm curve of my motors, and I don't have good data on how the airfoils will perform. So the next step is to order up some more props with various angles of attack, and to see how they perform.