\section{The zero endpoint and type \texorpdfstring{$\widetilde B_n$}{B-tilde}} \label{ab:section} Fix $n\geq3$. Let $W^0$ be the affine Coxeter group of type $\widetilde C_n$, with generators $s_0,\ldots,s_n$ in path order. Let $H^0$ be its Hecke algebra with weight zero at $s_0$, weight $b$ at $s_n$, and weight $a$ at all other generators. This is the specialization $c=0$ of the $\widetilde C_n$ datum. Write $W$ and $H$ for the group and algebra of type $\widetilde B_n$. The fork generators $t_0,t_1$ are adjacent to $t_2$, followed by the chain $t_2,\ldots,t_n$ with an edge of order four at the last pair. The weights are $a$ at $t_0,\ldots,t_{n-1}$ and $b$ at $t_n$. A superscript $0$ distinguishes the standard basis and the coefficient-of-identity trace of $H^0$. \begin{lemma} \label{ab:crossed} The assignments \[ t_0=s_0s_1s_0,\qquad t_i=s_i\quad(1\leq i\leq n). \] embed $W$ in $W^0$ as an index-two subgroup. Let $\gamma$ be the diagram automorphism interchanging $t_0$ and $t_1$ and fixing $t_2,\ldots,t_n$, acting on $H$ by permuting the normalized generators. Setting $U=N^0_{s_0}$ gives \[ H^0=H\rtimes\langle U\rangle,\qquad U^2=1,\qquad UhU=\gamma(h)\quad(h\in H). \] Under this embedding, $N_x=N_x^0$ and $N_xU=N_{xs_0}^0$ for $x\in W$. In particular, $\tau^0|_H=\tau$ and $\tau^0(HU)=0$. \end{lemma} \begin{proof} On $\mathbb R^n$ take the $\widetilde C_n$ alcove $1/2>x_1>\cdots>x_n>0$. Its walls give $s_0$ at $x_1=1/2$, $s_i$ at $x_i=x_{i+1}$ for $1\leq i0$ and sends the intersection for $k=0$ to the other standard fork facet. The inclusion at each intersection preserves the standard trace. For $k>0$ the finite algebra of $F_k$ is a crossed product of the intersection algebra by $\langle U\rangle$. Let $\sigma$ be a simple type of the larger algebra and $\eta$ a simple type in its restriction. Over a splitting field of characteristic zero, Clifford theory gives one simple $\sigma$ above an orbit of two inequivalent types $\eta$ and $\gamma(\eta)$, with both restriction multiplicities one. Comparing the coefficient of the character of $\eta$ in the restricted trace gives $s_\sigma^{-1}=s_\eta^{-1}$. For an invariant $\eta$ there are two extensions, each of restriction multiplicity one. Their reciprocal Schur elements sum to $s_\eta^{-1}$. They are equal because the quotient sign interchanges the extensions and preserves the trace. Thus $s_\sigma=2s_\eta$ in this case. The algebras agree for $k=0$, so in every case the finite spectral degree is preserved. Choose $\sigma$ to be an attaining constituent of $\pi$ on $F_k$. At least one constituent of its restriction to $F_k\cap W$ occurs in the restriction of $\rho$ or of $\gamma(\rho)$. In the latter case conjugate this constituent and the intersection by $\gamma$. This gives a constituent on a standard finite parabolic of $H$ in the restriction of the prescribed $\rho$. Its spectral degree is $A_R$ in either case. For integral $a,b$, this degree is integral by the type $B$ symbol formula for the original constituent on $F_k$, with endpoint weights $0,b$. \end{proof} The $\widetilde B_n$ application of Proposition~\ref{prop:leading} uses the induction tori and intertwiners for the reduced datum $R_0=C_n$, $X=Q(C_n)$. The stabilizer commutant is described by \cite[Theorem~5.5]{DelormeOpdam}, and \cite[Proposition~6.8]{DelormeOpdam} gives the groupoid comparison for this datum. The projected semisimple lattice can be seen by subtracting each type-$A$ block's mean from its coordinates and retaining the tail coordinates. Applied to $\mathbb Z^n$ this gives the product of the type-$A$ weight lattices and the weight lattice $\mathbb Z^\ell$ of the type-$C_\ell$ tail, where $\ell$ is the tail size. Lifts to $\mathbb Z^n$ of a point of this product differ by adding an integer constant to each type-$A$ block. If a block size is odd, one such shift changes the total parity and every point has a lift to $Q(C_n)$. If all block sizes are even, parity is well defined on the product lattice and the projected lattice from $Q(C_n)$ is its index-two kernel. When there is at least one type-$A$ block, the order-two character of this product lattice trivial on the kernel is nontrivial on an even-sized block. On that block it is the nontrivial quadratic twist of the Steinberg torsion label, which acts freely on the set of such labels. The resulting groupoid comparison is precisely that of \cite[Proposition~6.8]{DelormeOpdam}. For the full Levi there are no type-$A$ blocks, the unramified torus is a point, and the stabilizer description is the discrete-series block itself. In every case the compact model and the generalized Bernstein weights use the actual projected lattice and the unramified quotient of $Q(C_n)$. The normalized relative density on each torus has a Laurent polynomial limit as used in Proposition~\ref{prop:leading}. In fact its root factors, after restricting to the torus, have the form $1-\zeta v^d z^\nu$ and their conjugates, where $z^\nu$ is a torus character, $d$ is a real exponent and $|\zeta|=1$. Nonstationary factors have uniform normalized limits. A stationary denominator is a product of primitive Laurent binomials. Real analyticity of the exact density on the compact torus cancels each such factor because the compact points of every component of its divisor are Zariski dense \cite[Theorem~3.25 and Proposition~3.27]{Opdam}. After cancellation the normalized density converges uniformly. The proof of Proposition~\ref{prop:leading}, using the finite Bernstein expansion and the actual stabilizer action, now gives the full equivariant Laurent polynomial range for $H$. In applying its central cutoff construction one chooses the integral multiples of the dual real exponents in $Q(C_n)$. This is a full-rank lattice, so the choice is possible. Proposition~\ref{ab:facet} gives a same-degree finite source in every unitary stabilizer block. Hence the ordinary ideal generation and the canonical filtration in Proposition~\ref{prop:kl-filtration} apply to $H$. The trace used for the signed-unit argument can be computed by the Plancherel formula of $H^0$. To see its compatibility with the leading product of $H$, restrict all the compact families of $H^0$ to $H$. The spectral level of every constituent is the level of the family by Proposition~\ref{ab:facet}. For a compact family $\pi^0_{\alpha,v,z}$ of $H^0$ write $\mu^0_{\alpha,v}$ for its Plancherel measure on the family torus. Lemma~\ref{ab:crossed} gives the exact identity \[ \tau(h)=\sum_\alpha\int \operatorname{Tr}(\pi^0_{\alpha,v,z}(h))\, d\mu^0_{\alpha,v}(z)\qquad (h\in H). \] After normalizing the matrices by $v^{-A_R}$ and the measures by $v^{2A_R}$, its two-factor limit gives an isometry on the orthonormal basis indexed by $W$. The integral spectral projections in Proposition~\ref{prop:leading} act on each restricted fiber by the corresponding scalar cutoff. Indeed the central Bernstein elements act on each fiber by its central character on $H$, and the polynomial iteration defining the projections then selects the real exponent orbit. We now identify the product under the isometry. For fixed $x,y\in W$ the leading product column is the constant term of the normalized multiplication of $N_x$ on $N_y$ from Proposition~\ref{prop:leading}. The three-factor Plancherel formula identifies its image with the orthogonal projection of the pointwise product onto the closed image of the isometry. The norm of the column converges in the standard basis Hilbert space by the operator-norm convergence of that normalized multiplication. The four-factor Plancherel formula identifies the limit of the squared norms with the squared norm of the pointwise product itself. Equality of these norms places the product in the image. For the unit, apply \(P_+(v)\) from Proposition~\ref{prop:leading} to \(N_1\). In the restricted families this gives \(v^{A_R}\) times the identity matrix. Its squared norm tends to the sum over the families of the integrals of their dimensions against the limiting measures. Pairing with each \(N_w\) shows that the image of its constant coefficient is the orthogonal projection of the family identities onto the closed image of the isometry. Since the squared norms agree, this projection is the family identity itself.