Preprint
Article

This version is not peer-reviewed.

The Lonely Runner Problem on Pair-Sum Lattices: A Lossless Finitization and a Conditional Scaling-Closure Theorem for the Covering Obstruction

Submitted:

14 November 2026

Posted:

18 November 2026

You are already at the latest version

Abstract
We prove an exact identity reducing the maximum loneliness of a Lonely Runner instance to a finite maximum over cyclic groups: for every speed vector V of n positive integers, M (V) = max{u,w} τ (u, w), where τ (u, w) is the optimal loneliness on the pair-sum lattice of denominator N = vu + vw. The analytic core is folklore; the contributions are the sharpening that sum lattices alone suffice, the exact equality, and the lossless reformulation of the Lonely Runner Conjecture as the finite statement (TAU-n). This is a per-instance finitization—for each fixed V it replaces the defining supremum by a finite maximum over \( \binom{n}{2} \) pair-sum lattices, but the denominators N = vu + vw grow with V and no uniform bound on N is claimed. The reformulation turns the conjecture into a family of covering problems, and we develop that covering geometry: an exact size law, a class-group reduction at prime moduli, a lift law, and complete classifications of the unit coverings at the prime moduli 7, 11, 13 (three bad sets) and 7, 13, 17, 19, 37 (four bad sets). These five primes form the entire unit-capable rigid zone at n = 5 within the verified range—a striking but unproved finiteness statement. Empirically, on the full corpus of 3,712,576 primitive speed sets with n = 5 and speeds ≤ 56, proved lemmas close 75.6% of instances, per-modulus verification closes all of them, and the residual is localized to a single named class (mixed order signatures at composite moduli); the proved-coverage rate declines as the corpus grows (93.4% → 78.2% over V = 24 → 56) and its asymptotics are unresolved. The paper’s culminating result is a conditional theorem: under three named open hypotheses—the multi-fiber survivor inequality Hq± (subsuming the dilate-sparsity input H2m), the distinct-family volume law H1g, and the resonance budget H2r—the fiber cascade closes the covering obstruction at the frontier cells (1K, T −3U ) of every rung T ≥ 8 at N = p2 within a depth that is OT (log k) (no k-uniform linear bound is claimed). The reduction chain and every other input are proved or machine-validated, and each constant’s provenance (derived, measured, or fitted) is disclosed in the hypothesis ledger. That proved/hypothesized boundary has a structural type, verified statement-by-statement: every proved result is a closure, invariance, exhaustiveness, or machine-verification statement, while every remaining obligation is a uniform growth-in-k margin—a type no result of this paper proves. The one instrument shaped to produce that type, a transfer-matrix model of the strand census, closes by theorem in its single-anchor scope: its constrained transfer matrices are partial permutations of spectral radius exactly 0 or 1, and the missing margin is provably external to that lossless dynamics. This is a structural paper: it does not solve or reduce the Lonely Runner Conjecture, proves no new cases of it, and claims nothing beyond the verified ranges stated; the three open hypotheses are not silently established anywhere.
Keywords: 
;  ;  ;  ;  ;  ;  
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.