The paper is about a subtle but important error in how a mathematical formula gets passed between research papers. The original formula, due to Spencer Bloch, connects two different ways of computing a quantity called a regulator on an elliptic curve: one side involves integrating a certain differential form along a path on the curve, and the other involves a special function called the elliptic dilogarithm applied to a "diamond product" of points. The formula works correctly, but the question is exactly which path you integrate along, and what numerical factor appears in front.
When an elliptic curve defined over the rational numbers has only one connected piece when you look at its real points, the natural path that researchers in the Mahler measure literature have been using (the generator of a subgroup called the anti-invariant part of the first homology) turns out to be twice as long, in an appropriate sense, as the path that correctly normalizes Bloch's formula. This means that some papers in the area have been writing the formula with an implicit factor of 2 missing, which could lead to errors in numerical or exact computations. The authors trace this discrepancy to a simple index calculation using an earlier result by Brunault and verify it concretely on two specific elliptic curves, checking the numerical agreement to high precision.
The practical upshot is that some published results using this transmitted form of the formula need to be checked for this normalization issue. The authors look at specific theorems by Touafek and Bertin and find that most of their conclusions survive intact, either because they happen to involve the case where the curve has two real components (where no discrepancy arises) or because the numerical values are still consistent with the corrected formula. However, the derivation in one of Touafek's results contains a gap related to exactly this normalization point. The authors are careful to flag that a full audit of the literature using this formula is a larger task they leave open.