Browsing through studies, it seems that, other than at extremes, crank length doesn't make a significant difference to power.

That is the conclusion of the Martin study quoted above. Determinants of maximal cycling power: crank length, pedaling rate and pedal speed. - PubMed - NCBI

And also two more studies:

Effect of crank length on joint-specific power during maximal cycling. - PubMed - NCBI

"These data demonstrate that crank length does not affect relative joint-specific power once the effects of pedaling rate and pedal speed are accounted for. Our results thereby substantiate previous findings that crank length per se is not an important determinant of maximum cycling power production."

Effects of crank length on maximal cycling power and optimal pedaling rate of boys aged 8-11 years. - PubMed - NCBI

"Power produced when using the 170 mm cranks [mean (SEM)] [364 (18) W] did not differ from that produced with the LL20 cranks [366 (19)]. Optimal pedaling rate was significantly greater for the LL20 cranks [129 (4) rpm] than for the 170 mm cranks [114 (4) rpm]."

However, it seems there are some potential differences:
- Shorter cranks require higher cadence to get the same power as longer cranks. So if you prefer faster or slower cadence, you might prefer shorter or longer cranks.
- Shorter cranks require less range of motion at the leg joints. So if you already have limited ROM/flexibility, or conversely if you are worried about losing ROM/flexibility (use it or lose it), you might prefer shorter or longer cranks.
- Shorter cranks keep your knees lower at the top of the pedal stroke, which could permit a flatter back, so depending on flexibility and tibia/femur length, and on your need to ride in an aero position (TT or climber?), you might prefer shorter cranks.
- Shorter cranks require lower gearing, so depending on whether your lowest gearing needs to be and/or can be made lower (like, are you already using the lowest gearing your bike can have, to climb), you might not prefer shorter cranks.
- Shorter cranks require a higher saddle, which may result in a more rearward saddle, and possibly some setback adjustment.

According to the 20% of leg length formula that some propose, I should ride 165 mm cranks. I'll keep an eye out for cranks in that length (that I'd want to see on my bike . . . )

Come to think of it, I've previously had occasion to drill holes in a solid alloy crankarm, and it was pretty easy using a drill press. I suppose I could buy a cheapish crank, drill and tap a bunch of holes, and experiment away, without spending several hundred dollars on "adjustable" cranks that weigh 1000 g (PowerCranks etc).