The math is much simpler

 Yglesias mentions the math of economics.

Treat time as a variable of uncertainty, and output of the model.  What is left, People bunching up, too much or two little 

The existing model assumes that a third entity can buy and sell future time to make the world synchronous.  Yes, but they cannot be completely accurate. The only thing the mechanics can adjust is by relative shifting of the congestion, locally.

It is not the length of time,'eventually it a;ll averages out'.  No, eventually 'it all bounces around a few optimum configurations'.  Each eventually become more accurate when N goes to its optimum number, for all the finite set of eveuntualies.  That is as far as we can predict..

The Ygflesias statement is that under current conditions, the world is best stabilized by 6-9 great population masses, one billion each. But his equilibrium assumes we have indefinite time, we do not, the humans timeout.  If we remove the infinity, the better estimate is the number of well rounded cities.  Likely cities capture the bulk of transaction endogenously.  Take a city of N people making transactions. Model thgem as savers and borrowers each, of some proportion.   You find a hard limit to the density network in a city, so that proper proportion will have larger uncertainty if you jam too many transaction. Or if you have too few transactions.  

Go ask the retailer, how accurate are deliveries. They will tell you about 1/5 accurate.   The jitter.  Unless we move to the matrix model, that jitter is a hard bound, currently.  We can go to automation, but we still support people jitter, and that jitter is who we are. A hard bound. So we really do not know life better than 3-4 bit accuracy.  Your plans for a billion at the balance points need to be decomposed into a hierarchhy of 4 bit measurements.