\section{An effective period without total unimodularity}\label{sec:effective} We prove Theorem~\ref{thm:effective} while retaining integer feasibility. The dimension depends on the fixed order $t$, not on $B$. \subsection{A finite list of elements and lengths} Let $d=\varphi(t)$, $m=2^{w-1}$, $E=2^w-1$, and \[ \Delta_t=\max\{|\det R_t[I,J]|:\det R_t[I,J]\ne0\},\qquad L_t=\lcm\{|\det R_t[I,J]|:\det R_t[I,J]\ne0\}, \] including the empty minor 1. Fix a residue $j$ modulo $\rho=\rho_t$, so that the lattice $\Gamma_j$ and first gauge $\mu=\mu_t(j)$ are constant. The choice \eqref{eq:rho} ensures that $\mu\rho$ is integral and that the coordinate denominators of $\Gamma_j$ divide $\rho$. The upper bound in Theorem~\ref{thm:slopes} and \eqref{eq:maximum-principle} give, for a minimum false positive, \begin{equation}\label{eq:beta-bound} \|\beta\|_{Z_t}\le \mu+\frac{4\mu+1}{D}+\frac{4\mu+E}{D^2},\qquad D=B-1. \end{equation} Indeed $\lceil K_t(B)/t\rceil\le\lceil U_t(B)/D\rceil$. Choose an integer $D_0$ satisfying \begin{equation}\label{eq:D0} D_0\ge4\mu+E,\qquad D_0\ge\frac{2(4\mu+1)}{\mu},\qquad D_0^2\ge\frac{2(4\mu+E)}{\mu}. \end{equation} For $D\ge D_0$ the gauge is at most $2\mu$, and each CRT coordinate has absolute value at most $2m\mu\le2$. Thus the finite list \begin{equation}\label{eq:finite-beta} \mathcal S_j=(\Gamma_j\cap[-2,2]^d)\setminus\{0\} \end{equation} suffices. Enumerate the grid with denominator $\delta_j$ and test lattice membership: on a ramified branch require $(1-e\zeta)\beta\in A$. The full grid is not substituted for $\Gamma_j$. Writing $B=\rho n+j$, the minimum length has the form \begin{equation}\label{eq:finite-offset} k=t\mu\rho n+c,\quad \left\lceil t\mu j-t(\mu+1)+1\right\rceil \le c\le\left\lfloor t\mu j+t(3\mu+2)\right\rfloor . \end{equation} The lower bound follows from \eqref{eq:slope-bounds}. For the upper bound use $U_t(B)/D\le\mu B+3\mu+(4\mu+E)/D$ and $D\ge4\mu+E$. This offset interval is finite and independent of $n$. With $a=c\bmod t$ chosen in $\{0,\ldots,t-1\}$, we have $L=\mu\rho n+\lfloor c/t\rfloor$. \subsection{The fibre as a monoid} For $\beta\in\mathcal S_j$ and an offset $c$, define \[ \alpha(n)=H_{\rho n+j}\beta,\qquad v_r^{\rm cap}(n)=(\rho n+j-1) \bigl(\mu\rho n+\lfloor c/t\rfloor+\mathbf1_{r0$, then $z_i\ge\Delta\ge q_i$; if $q_i\le0$, that coordinate does not decrease. All coordinates remain nonnegative, contradicting lexicographic minimality. The set $I$ of large coordinates and the vector of the remaining coordinates give \eqref{eq:monoid-union}. The opposite inclusion is immediate. \end{proof} Decomposition into translates of monoids with independent generators is classical; see Chistikov--Haase~\cite[Proposition 5]{CH}. The proof above gives a minor-based bound under the height-one hypothesis needed for $M_t$. It assumes neither normality nor absence of integer holes, and is not a new general theory of semilinear sets. \subsection{Congruences, intercepts, and cutoffs} In a component $(I,v)$, choose rows $J$ such that $M_t[J,I]$ is square and invertible, and put $d_I=|\det M_t[J,I]|$. Cramer's rule gives the only possible component coordinates: \[ z_i(n)=N_i(n)/d_I, \] where the $N_i$ are integer polynomials of degree at most two, absorbing the determinant's sign into the numerators. The exact conditions are \begin{align} M_IN(n)&=d_I(Y(n)-M_tv),\label{eq:span-test}\\ N_i(n)&\ge0,\label{eq:sign-test}\\ N_i(n)&\equiv0\pmod{d_I}.\label{eq:congruence-test} \end{align} For $I=\varnothing$, take $d_I=1$ and require only $Y=M_tv$. Every nonzero polynomial residual in \eqref{eq:span-test} has finitely many roots. Every nonzero polynomial in \eqref{eq:sign-test} eventually has the sign of its leading coefficient. Conditions \eqref{eq:congruence-test} have period $d_I$ in $n$, and $d_I$ divides $L_t$. For $0\le h