The covering radius of a shift space measures the worst-case delay in synchronization: given a channel defined by a constrained sequence space, how long must you wait (at worst) before you can identify the current state from the observed output? Information-theoretically, it quantifies the maximum ambiguity of the channel.
Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2603.21449 (Meyerovitch and Young). The essay defines the covering radius as a worst-case synchronization delay. The paper's definition is different: the covering radius of a code is the smallest R such that every word is within Hamming distance R of some codeword, and the covering radius of a shift space is the analog for infinite sequences, relevant to transmission over noisy channels. The word synchronization does not appear in the paper. The result is as the essay states it (rational for primitive sofic shifts, with an algorithm from a labeled graph presentation), but the essay's explanation of why, that rational eigenvalues of the adjacency matrix produce rational radii and that finiteness forces rationality, is the essay's own; the paper's introduction presents rationality as a question raised by earlier ad hoc computations in which the answer happened to be rational.
The paper on the covering radius of sofic shifts (arXiv: 2603.21449) proves that for primitive sofic shifts, this quantity is always a rational number, and describes an algorithm to compute it from a labeled graph presentation.
Sofic shifts are the class of shift spaces recognizable by finite automata — they include shifts of finite type and their factors. Primitivity means the underlying graph is strongly connected and aperiodic. The covering radius being rational is not obvious: it's defined as a limit involving worst-case distances, and there's no a priori reason for the answer to be algebraic, let alone rational.
The rationality comes from the graph structure. The worst-case synchronization delay is determined by paths in the labeled graph, and the combinatorics of these paths — lengths, weights, overlaps — are governed by the graph's adjacency matrix. Rational eigenvalues produce rational covering radii.
The through-claim: the worst-case delay in a finite-state channel is rational because the channel is finite. The covering radius feels like an analytic quantity — a limit, a supremum over all possible inputs. But because the shift space is defined by a finite labeled graph, the extremal behavior is achieved by finite combinatorial objects, and the answer is rational. Finiteness of the description forces rationality of the invariant.
2603.21449. Information theory / sofic shifts / covering radius / symbolic dynamics / labeled graphs.