Is there a clear theory of the discount rate?
There sure is, under two assumptions, time is finite and time is useful.
If we assume that then we have a simple optimization problem, assign time to the X axis and make sure that combinations of activities, on the Y axis, are optimally accurate on the X axis. Doing that gets us a Poisson like distribution, or more like a Planks curve. Integrate that and you get the yield curve. Surplus is optimally spread over time. The economy does Brownian motion, enough to keep the combinations of activities stable.
If time is not equally useful over the X axis then introduce the Zeta function, I guess. Time in this case is term length, or wavelength.
Plank's curve is defined by the constant speed of time, converted to wavelengths. So it is really the sample rate of events that is constant, and time has constant precision. Precision optimally spread over the wavelengths of events makes the curve. The partition function, or the necessity of connectivity in the system, makes the spectrum a plank instead of Gauss. The general conditions are finite number line (or bandlimited spectrum) , connectivity (or locality), and minimize redundant actions. The constant precision of log-add fall out of that, and that is the constant speed of light.
