On a hereditarily precompact quasi-metric space, every non-contractive bijection $F$ is asymptotically an isometry along its backward orbits: for all $a$ and $\varepsilon>0$, $0\le d(Fx,Fy)-d(x,y)<\varepsilon$ for $x$ far in the past of $a$ and all $y$. Zavarzina asked whether every such bijection of a precompact space is an isometry; the author's toll spiral (arXiv:2610.13973) answered no. Each such $F$ has a register, a quasi-pseudo-metric $λ^-\le d$ for which $F$ is an isometry, with round-trip distances totally bounded on orbits; after completion, $F$ acts on each orbit as a rotation of a compact monothetic group, its rigid shadow. The defect $d-λ^-$ from $x$ to $y$ is at most the fee of $x$ (the supremum of its defects to earlier iterates) and at least the excess of $x$'s fee over $y$'s now or at any earlier time, less their round-trip register distance; the fee never decreases forward along orbits and tends to zero backward. In this sense, every such $F$ is a toll over its rigid shadow. The fees bound the defect but do not determine it: three spirals show it can depend on $y$, vanish at a distant $y$ even when $x$ has a positive fee, and be positive toward later iterates of $x$. The toll-spiral paper asked whether every such map is conjugate to a tolled extension, a fixed toll zone over an isometric shadow, possibly with infinitely many zones. If only descents are charged, the answer is no, even up to bi-Lipschitz conjugacy; on each infinite orbit, a map is such an extension over its register exactly when its ascents carry no defect. Fees may vanish at any rate, but for zone tolls (two-valued, shift-monotone, subadditive along their routes) they reach zero at a finite depth. The general results hold on quasi-pseudo-metric spaces where every sequence has a left K-Cauchy subsequence; all but the statements listed at the end are machine-checked in Lean.