On any given day, more people fly from New York to Miami than from Miami to New York. The flow is asymmetric — driven by events, seasons, economics. But zoom out. Over months, then years, the asymmetry fades. The flows approach balance. Not because anyone is coordinating the balance. Because the system is diffusive at long timescales.
Five years of intercity mobility data (arXiv:2603.21552) show that over half of all city pairs converge toward effective flow balance under temporal coarse-graining, with the normalized directional imbalance decaying as a power law in the averaging window. The system transitions from broken to restored flow symmetry as you zoom out.
Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2603.21552. The findings match the essay. The data, though, are five years (2021 to 2025) of daily intercity flows from the Baidu Qianxi project, inferred from mobile phone geolocation, covering prefecture-level cities in China. The essay's New York and Miami flights are an illustration of the idea and not the data, and the paper is not about air travel. The essay's remarks that the microscopic dynamics are genuinely irreversible and that neither description is fundamental are the essay's own.
The mechanism is stochastic: mobility decomposes into a directional drift (the consistent asymmetry) and correlated fluctuations (the noise). At short timescales, drift dominates. At long timescales, the fluctuations — which don't have a preferred direction — wash out the drift. The variance scales near-diffusively, which is why the power law emerges.
Three regimes coexist: flow-balanced pairs (the majority), persistent drift-dominated currents (a few routes that stay asymmetric no matter how long you average), and transitional cases. The classification depends on the drift-to-fluctuation ratio for each city pair.
The structural insight: a driven, far-from-equilibrium system spontaneously produces equilibrium-like statistics at coarse resolution. This isn't approximate equilibrium — the microscopic dynamics are genuinely irreversible. The detailed balance is emergent at a macroscopic scale, not inherited from a microscopic one.
The resolution changes the physics. At the daily scale, the system is out of equilibrium. At the annual scale, it looks thermal. Both descriptions are correct. Neither is fundamental. The timescale of observation determines which symmetry you see.
Note (2026-11-11): read in full, after the earlier note, against arXiv 2603.21552. Three points. (1) The essay says persistent drift-dominated currents are 'a few routes'. In the paper, 21 percent of the city pairs are drift-dominated, which it calls 'a substantial minority', alongside the roughly half that converge, a crossover group, and under 5 percent atypical pairs. (2) The essay says the classification 'depends on the drift-to-fluctuation ratio'; the paper sorts pairs by the fitted slope of the imbalance against the averaging window at small and large windows (a pair is fluctuation-dominated if both slopes are below minus 0.1, drift-dominated if both are within 0.1 of zero), and fits a stochastic model of drift and correlated fluctuations afterwards. The decay exponents average about minus 0.47, near the minus 0.5 expected for uncorrelated fluctuations, and the paper describes the variance scaling as near-diffusive with regime-dependent deviations. (3) The paper uses 'detailed balance' in an effective, flow-level sense, and reports that the three-regime structure and the exponents hold when only the post-COVID years (2023 to 2025) are used. The essay's closing claims about symmetry and about which description is fundamental are its own.