I don't see an obligatory train wreck either. But you probably have the data already. Power and cadence gives you torque (average per rotation) with a little bit of post-processing, from there a few exploratory scatter plots will tell you if there is a pattern indicating a relationship.
Ramp test data tells me that as power demand increases, I choose to use a higher cadence, I do upshift as speed increases but not enough to keep cadence constant.
With a cubic power curve, 26% faster = doubling the power. So on a constant gear, 26% increase in cadence 59% increase in torque. Constant cadence, 100% increase in torque. Constant torque, 100% increase in cadence therefore 59% reduction in gearing for 26% higher speed. So I'm certainly choosing to reduce the increase in torque that would have been required for a constant cadence, but definitely not keeping it constant.
I could check on two turbotrainers with completely different power curves to see what consistency I exhibit, if any.

Am I making the right choices? Who could tell me, apart from my own internal feedback?
My first guesses are that I have chosen a crank length that provides the "right" balance between pedal force and limb articulation, and then choose a cadence that provides the "right" balance between pedal force and speed. A shorter crank and lower gear could provide unchanged pedal force and speed for a given power, it would be produced with a higher cadence and reduced limb articulation. What is most effective for (a) an individual performance and (b) positive adaptation for individual future performances are further questions that I lack the knowledge to answer. I probably couldn't even fathom the correct intermediate questions.