\section{Reaction-count aliases determine the first loss term}\label{sec:alias} Let $\nu_1,\ldots,\nu_r\in\mathbb Z^d\setminus\{0\}$ be a finite mark set, with $p_i>0$, $\sum_i p_i=1$. Repeated marks are allowed. Write $S$ for the matrix with these columns and let \begin{equation}\label{eq:frozen} Z_i\stackrel{\rm ind}{\sim}\Pois(\lambda p_i),\quad K=\sum_i Z_i,\quad D=SZ,\quad F_p(\lambda)=1-\lambda^{-1}\E\Var(K\mid D). \end{equation} This is the frozen-intensity experiment; it is not automatically the exact one-bin experiment of a state-dependent network. \begin{definition}[Unequal-length alias depth] For $d\in\mathbb Z^d$ and $k\ge0$ put \begin{equation}\label{eq:bk} b_{d,k}=\sum_{\substack{z\in\NN^r:\,Sz=d,\,|z|=k}} \prod_{i=1}^r\frac{p_i^{z_i}}{z_i!}. \end{equation} If the fibre over $d$ supports at least two distinct lengths, denote its smallest and next smallest lengths by $m_d$ and $k_d$. Let $h=\min_d k_d$, or $h=\infty$ if no such fibre exists. \end{definition} Equivalently, $h$ is the smallest value of $\max\{|z|,|z'|\}$ for $Sz=Sz'$ and $|z|\ne|z'|$. $\NN$ denotes the nonnegative integers. Same-length aliases do not hide the total reaction count and hence do not lose information about its common intensity. \begin{theorem}[Alias-depth expansion]\label{thm:alias} Let $F_p$ be the retained fraction of the frozen experiment \eqref{eq:frozen}, and let $h$, $b_{d,k}$, $m_d$, $k_d$ be as in the preceding definition. If $h<\infty$, then $h\ge2$ and \begin{equation}\label{eq:alias} 1-F_p(\lambda)=C_p\lambda^{h-1}+O_p(\lambda^h), \qquad C_p=\sum_{d:\,k_d=h}(h-m_d)^2b_{d,h}>0. \end{equation} If $h=\infty$, then $F_p(\lambda)=1$ for every $\lambda>0$. \end{theorem} The proof of Theorem~\ref{thm:alias} is in Appendix~\ref{app:alias-details}. The coefficient has a random-walk form. Let $\xi_1,\xi_2,\ldots$ be independent with $\PP(\xi_\ell=\nu)=\sum_{i:\,\nu_i=\nu}p_i$, put $\zeta_k=\xi_1+\cdots+\xi_k$ ($\zeta_0=0$), and let $\mathcal Z_k=\{d:\PP(\zeta_k=d)>0\}$. \begin{corollary}[Random-walk form of the alias coefficient]\label{cor:alias-rw} If $h<\infty$, then \begin{equation}\label{eq:alias-rw} C_p=\frac1{h!}\sum_{m=0}^{h-1}(h-m)^2\,\PP(\zeta_h\in\mathcal Z_m). \end{equation} For every finite nonzero mark law, $1-F_p(\lambda)=\bigl[2\PP(\zeta_2=0)+\tfrac12\PP(\zeta_2\in\mathcal Z_1)\bigr]\lambda +O_p(\lambda^2)$, and the bracket vanishes exactly when $h\ge3$. In particular: \begin{enumerate} \item for nearest-neighbour marks $\pm e_j$ in $\mathbb Z^d$ with probabilities $p_j^\pm\ge0$ (marks of probability zero omitted), one has $h=2$ if some $p_j^+p_j^->0$, and then $C_p=4\sum_jp_j^+p_j^-$; otherwise $h=\infty$; \item for the simplex marks $e_1,\ldots,e_d,-(e_1+\cdots+e_d)$ in $\mathbb Z^d$, $d\ge1$, with probabilities $p_1,\ldots,p_d,p_0>0$, one has $h=d+1$ and $C_p=(d+1)^2\prod_{i=0}^dp_i$. \end{enumerate} \end{corollary} \begin{proof} By the multinomial theorem, $b_{d,k}=\PP(\zeta_k=d)/k!$. For $m