\section{Resonant amplification and candidate channel mechanisms} \label{cx:section} We examine several mechanisms relevant to the fixed-section channel question. A passive device can have large internal fields; locally evanescent fields can carry positive flux through interference; and a finite-dimensional resonator model does not by itself realize a Maxwell device. The examples distinguish exact Maxwell constructions from finite-dimensional models, and do not identify a model port with a physical channel through the prescribed section. The physical examples use the fixed real scalar material class $1\leq\varepsilon_r\leq\varepsilon_+$ at one positive frequency, with vacuum exterior. The section and thresholds are fixed, whereas support size and resonance lifetime may grow. The radial, local-flux, and mirror-network examples originate in the prior report \cite{priorreport}; they are rederived here with their hypotheses explicit. The physical counterexample to the quantified fixed-section assertion is given separately in Theorem~\ref{uc:counterexample}. \subsection{An exact bounded-index Maxwell amplification obstruction} First consider a single incoming spherical wave surrounded by concentric dielectric layers. Adding sufficiently many layer pairs makes the core field arbitrarily large while the outgoing wave still carries exactly the incident power. This provides a concrete reason that passivity alone cannot supply the uniform local field bound needed by Section~\ref{ch:section}. It is a one-channel reflection example, so it does not itself establish many channels through the prescribed square. \begin{proposition}[Unbounded local amplification, one reflected channel] \label{cx:bragg} For every fixed $\varepsilon_+>1$ and $k_0>0$ there are a fixed ball $B_{r_0}$ and finite radial dielectric structures $\varepsilon_M$, with $1\leq\varepsilon_M\leq\varepsilon_+$, for which a unit-power incoming spherical radiation channel produces electric and magnetic squared field norms in $B_{r_0}$ growing proportionally to $q^{2M}$, for some fixed $q>1$. The exterior radius grows linearly with $M$. The selected scattering coefficient has modulus one and the net power through every concentric sphere is zero. \end{proposition} \begin{proof} Choose $11$ its sign alternates, since its value is $(-1)^m(1-x^{-2})$. Choose such a root $x_0>1$ large enough that $q^2+2/x_0^2\leq\varepsilon_+$, and put $r_0=x_0/k_0$. The root is not simultaneously a zero of $\widehat j_1$, by uniqueness for its second-order ordinary differential equation away from zero. For a tangential, surface-divergence-free vector spherical harmonic $X_{1m}$ with $\int_{S^2}|X_{1m}|^2\,d\Omega=1$, set \begin{equation} E=\frac{u(r)}r X_{1m},\qquad H=\frac1{i\omega\mu_0}\nabla\times E, \qquad u''+\left(k_0^2\varepsilon_M(r)-\frac2{r^2}\right)u=0. \label{cx:radial-maxwell} \end{equation} This is the exact TE multipole reduction of both Maxwell curl equations. At radial interfaces continuity of tangential $E$ and $H$ is exactly continuity of $u,u'$; permeability is constant and the normal electric component vanishes. The integrated outgoing power is \begin{equation} P(r)=\frac1{2\omega\mu_0}\operatorname{Im}(\overline u u'). \label{cx:radial-power} \end{equation} In the core take $\varepsilon_M=1$ and $u=C_M\widehat j_1(k_0r)$, which is regular at zero and has $u'(r_0)=0$. Append $M$ low/high shell pairs. A shell of type $n\in\{1,q\}$ has width $\pi/(2k_0n)$ and coefficient $\varepsilon_M(r)=n^2+2/(k_0r)^2$. Thus its radial equation is exactly $u''+(k_0n)^2u=0$. On the state $(u,u'/k_0)^T$, its transfer matrix is $\left(\begin{smallmatrix}0&1/n\\-n&0\end{smallmatrix}\right)$. A pair has transfer matrix $-\operatorname{diag}(q^{-1},q)$, so \begin{equation} R_M=r_0+\frac{M\pi}{2k_0}(1+q^{-1}),\quad u(R_M)=(-q^{-1})^M C_M\widehat j_1(x_0),\quad u'(R_M)=0. \label{cx:bragg-transfer} \end{equation} All shells obey the claimed material bounds. In the vacuum exterior put $u=a\widehat h^{(2)}_1(k_0r)+b_M\widehat h^{(1)}_1(k_0r)$, where $\widehat h^{(1)}_1(x)=-e^{ix}(1+i/x)$ and $\widehat h^{(2)}_1=\overline{\widehat h^{(1)}_1}$ on the positive real axis. Direct differentiation gives the Wronskian $2i$ and \begin{equation} |(\widehat h^{(1)}_1)'(x)|^2=1-x^{-2}+x^{-4},\quad b_M=-a\frac{(\widehat h^{(2)}_1)'(k_0R_M)} {(\widehat h^{(1)}_1)'(k_0R_M)},\quad u(R_M)=\frac{2ia}{(\widehat h^{(1)}_1)'(k_0R_M)}. \label{cx:bragg-exterior} \end{equation} Incoming power is $k_0|a|^2/(2\omega\mu_0)$; choose $|a|^2=2\omega\mu_0/k_0$. Consequently \begin{equation} |C_M|^2= \frac{4q^{2M}|a|^2}{|\widehat j_1(x_0)|^2 |(\widehat h^{(1)}_1)'(k_0R_M)|^2}. \label{cx:bragg-growth} \end{equation} For $x\geq1$ the derivative factor squared lies in $[3/4,1]$. The core fields are fixed nonzero profiles times $C_M$, proving the norm statement. A fixed interior section on which their traces are nonzero has the corresponding trace growth. Equation~\eqref{cx:bragg-exterior} gives $|b_M|=|a|$, while the core radial function is a constant times a real function. Equation~\eqref{cx:radial-power} and conservation then give zero total radial flux at every radius. No independently chosen trapped amplitude was inserted: $C_M$ is fixed by the incoming amplitude and the interface matching. \end{proof} \begin{remark}[Scope of the amplification example] \label{cx:bragg-scope} External passivity does not uniformly bound the incident-power-to-local-field map over the admitted class with unrestricted support. The prior report's $q=2$ example required $\varepsilon_+>4$; choosing $x_0$ and $q$ as above extends this obstruction to every $\varepsilon_+>1$. This does not disprove a uniform channel count, a fixed-support estimate, or a theorem restricted to counted fields with positive through-power. A complete spherical scattering phase is not an additional transmitted channel. This exact construction establishes neither an increasing number of transmitted directions nor the required absolute bypass bound. \end{remark} \subsection{Evanescent pairs can carry real power locally} An individual exponentially decaying or growing factor is often associated with reactive rather than transported power. A suitable phase relation between the two factors changes their interference: the combined field has positive real normal flux. The example below gives arbitrarily many such independent local solutions. Their existence in a collar does not show that they can all be generated by finite-power incident fields while meeting the global bypass requirement. \begin{proposition}[Infinite positive local flux index] \label{cx:evanescent} On a fixed rectangular section $D=(0,L)\times(0,W)$ and homogeneous source-free collar $D\times(-h,h)$, the space of smooth local Maxwell solutions has positive flux subspaces of arbitrarily large finite dimension. This remains true if every superposition has zero real normal flux on the four collar sidewalls. \end{proposition} \begin{proof} Let $k^2=\omega^2\mu_0\varepsilon_0\varepsilon_c$ and take every integer $m$ with $q_m=m\pi/L>k$. Put $\kappa_m=\sqrt{q_m^2-k^2}$ and \begin{equation} f_m(z)=a_m(e^{-\kappa_mz}+i e^{\kappa_mz}),\qquad |a_m|^2=\frac{2\omega\mu_0}{\kappa_mLW},\qquad E_m=(0,f_m\sin(q_mx),0), \label{cx:evanescent-electric} \end{equation} \begin{equation} H_m=\frac1{\omega\mu_0} (i f'_m\sin(q_mx),0,-i q_m f_m\cos(q_mx)). \label{cx:evanescent-magnetic} \end{equation} Both curl equations hold by $f_m''=(q_m^2-k^2)f_m$. The normal power density is \begin{equation} \frac{\operatorname{Im}(\overline f_m f'_m)\sin^2(q_mx)}{2\omega\mu_0} =\frac{\kappa_m|a_m|^2\sin^2(q_mx)}{\omega\mu_0}. \label{cx:evanescent-density} \end{equation} Sine orthogonality therefore gives, for every finite coefficient vector, \begin{equation} P_D\left(\sum_m c_m(E_m,H_m)\right)=\sum_m|c_m|^2. \label{cx:evanescent-gram} \end{equation} On the $x$ sidewalls $E$ vanishes, and on the $y$ sidewalls the normal Poynting component vanishes identically. Both statements hold for every finite superposition, not just for basis fields. \end{proof} \begin{remark}[Scope of the local evanescent example] \label{cx:evanescent-scope} A single decreasing exponential carries zero real normal power. Its coherent combination with an increasing exponential does not. Thus a propagating-wave count cannot be justified by discarding each evanescent term separately. These are exact \emph{local} vector Maxwell fields in ordinary material, but their norms at the collar ends grow rapidly with transverse order. No finite passive dielectric device producing these data from power-normalized external channels has been constructed. Zero sidewall flux is a property of the fields, not permission to install conducting walls. Nor does it control the field on the rest of the infinite section plane. Hence this proposition does not establish global absolute bypass or a physical counterexample. \end{remark} \subsection{An exact feedback diagnostic independent of mode reduction} \begin{proposition}[Lossless scattering algebra] \label{cx:mirror-network} For any $N$ and any $0<\xi_j\leq1$, reciprocal unitary two-port networks with isolated transmission powers $\xi_j$ can be cascaded with lossless phases to produce $N$ perfectly transmitted orthogonal network channels. \end{proposition} \begin{proof} For direction $j$, use two identical mirrors with $r_j=i\sqrt{1-\xi_j}$ and $t_j=\sqrt{\xi_j}$. The symmetric matrix $\left(\begin{smallmatrix}r_j&t_j\\t_j&r_j\end{smallmatrix}\right)$ is unitary. With propagation phase $p_j=i$, the convergent reflection series gives \begin{equation} t_{{\rm all},j}=\frac{t_j^2p_j}{1-r_j^2p_j^2}=i, \qquad r_{{\rm all},j}=r_j+ \frac{t_j^2r_jp_j^2}{1-r_j^2p_j^2}=0. \label{cx:mirror-feedback} \end{equation} The direct sum has transmission $iI_N$. The feedback inverse is exactly $\operatorname{diag}(\xi_j^{-1})$, compensating arbitrarily weak isolated links. \end{proof} \noindent\textbf{What is disproved.} An isolated-link singular-value tail together with external passivity does not control the completed scattering transmission. This proposition is stronger as an algebraic diagnostic than a coupled-mode approximation, but still supplies no bounded-index synthesis through the same physical $D$. \subsection{Matched resonator pairs: energy, bandwidth, and errors} The following model uses the energy/power normalization conventional in temporal coupled-mode theory \cite{fan2003,suh2004}. Its equations and all conclusions about them are proved directly; their agreement with a particular Maxwell device remains an additional question. \begin{definition}[MODEL: a reciprocal resonator pair] \label{cx:pair-model} Let $x,y$ be energy-normalized resonator amplitudes, so stored model energy is $|x|^2+|y|^2$. Let $a,b$ be incoming amplitudes with incident power $|a|^2+|b|^2$. For $J,\gamma>0$, set \begin{align} \dot x&=(-i\omega_0-\gamma)x-iJy+\sqrt{2\gamma}\,a, &a_-&=-a+\sqrt{2\gamma}\,x,\notag\\ \dot y&=(-i\omega_0-\gamma)y-iJx+\sqrt{2\gamma}\,b, &b_-&=-b+\sqrt{2\gamma}\,y. \label{cx:pair-dynamics} \end{align} The $-1$ direct reflection is part of the model's port convention. \end{definition} \begin{proposition}[Exact matching and cost in the model] \label{cx:pair-solution} For harmonic drive at $\omega=\omega_0+\Delta$ with $a=1,b=0$, define $D_\Delta=(\gamma-i\Delta)^2+J^2$. Then \begin{align} x&=\frac{\sqrt{2\gamma}(\gamma-i\Delta)}{D_\Delta},& y&=\frac{-i\sqrt{2\gamma}J}{D_\Delta},\notag\\ t&=\frac{-2i\gamma J}{D_\Delta},& r&=\frac{\gamma^2+\Delta^2-J^2}{D_\Delta}. \label{cx:pair-scattering} \end{align} The model is lossless and passive. At $\Delta=0$, $\gamma=J$ it has $t=-i,r=0$, internal energy $J^{-1}$, and left-to-right internal current $1$. At this matched setting, \begin{equation} |t(\Delta)|^2=\frac1{1+\Delta^4/(4J^4)},\qquad {\rm FWHM}=2\sqrt2 J. \label{cx:matched-linewidth} \end{equation} \end{proposition} \begin{proof} Substitution of harmonic time dependence gives the linear system with matrix $\left(\begin{smallmatrix}\gamma-i\Delta&iJ\\ iJ&\gamma-i\Delta\end{smallmatrix}\right)$; inversion proves \eqref{cx:pair-scattering}. Direct differentiation of the energy gives \begin{equation} \frac{d}{dt}(|x|^2+|y|^2) =|a|^2+|b|^2-|a_-|^2-|b_-|^2. \label{cx:pair-energy-balance} \end{equation} The internal left-to-right exchange is $I_{L\to R}=-2J\operatorname{Im}(\overline x y)$, from the individual energy derivatives. At matching $x=1/\sqrt{2J}$ and $y=-ix$, giving the energy and current. Finally, $|D_\Delta|^2=(2J^2-\Delta^2)^2+4J^2\Delta^2 =4J^4+\Delta^4$ gives the line shape and its half-power frequencies. The poles are at $\Delta=\pm J-i\gamma$, and hence remain strictly below the real axis for every member of the family. \end{proof} \begin{remark}[Model sensitivities] \label{cx:pair-sensitivity} At $\Delta=0$, writing $s=\gamma/J$ gives $|t|^2=4s^2/(1+s^2)^2$. Thus $|t|^2\geq\tau$ exactly when \begin{equation} \frac{1-\sqrt{1-\tau}}{\sqrt\tau}\leq s\leq \frac{1+\sqrt{1-\tau}}{\sqrt\tau}. \label{cx:rate-tolerance} \end{equation} At matching, a common frequency detuning obeys the threshold exactly when $|\Delta|\leq\sqrt2 J(\tau^{-1}-1)^{1/4}$. If both resonators instead have an additional unobserved amplitude decay $\kappa\geq0$, replace the diagonal rate in the equations by $\gamma+\kappa$ while keeping the driven/output couplings $\sqrt{2\gamma}$. For $\gamma=J$ and $\Delta=0$, \begin{equation} |t|^2=\frac4{[(1+\kappa/J)^2+1]^2}. \label{cx:loss-tolerance} \end{equation} Consequently fixed loss destroys the small-$J$ limit. To approach unit transmission it is necessary here that $\kappa/J\to0$. For a fixed threshold, the exact permitted ratio is $\kappa/J\leq\sqrt{2/\sqrt\tau-1}-1$. These costs cannot be suppressed by saying merely that each cavity has large $Q$: its leakage and errors must be compared to its weak link. \end{remark} \begin{theorem}[$N$ weak directions and a stable certificate in the model] \label{cx:n-model} Let $J=\operatorname{diag}(J_1,\ldots,J_N)$ with all $J_j>0$. At the common frequency $\omega_0$, take $N$ independent pairs with external rates $J_j$. Put \begin{equation} G=\operatorname{diag}(J^{1/2},J^{1/2}),\quad B_0=\begin{pmatrix}I&iI\\iI&I\end{pmatrix},\quad A_0=GB_0G,\quad E_L a=(a,0)^T,\quad E_R a=(0,a)^T. \label{cx:weighted-matrices} \end{equation} Their transmission and reflection are $T_0=-iI_N$, $R_0=0$; their internal energy matrix is $J^{-1}$ and exchange-current matrix is $I_N$. If the system matrix changes to $A=A_0+\mathcal E$, keeping the same port couplings, define $F=G^{-1}\mathcal E G^{-1}$ and $\rho=\|F\|$. Whenever $\rho<\sqrt2$, the system remains invertible and \begin{equation} \|T-T_0\|\leq\frac{\rho}{1-\rho/\sqrt2},\qquad T=2E_R^*(B_0+F)^{-1}E_L. \label{cx:weighted-error} \end{equation} In particular every input is transmitted with power greater than $\tau$ if \begin{equation} \rho<\frac{1-\sqrt\tau}{1+(1-\sqrt\tau)/\sqrt2}. \label{cx:rho-certificate} \end{equation} \end{theorem} \begin{proof} The unperturbed statement follows pairwise from Proposition~\ref{cx:pair-solution}. Also $B_0^*B_0=2I$, so $\|B_0^{-1}\|=1/\sqrt2$ and $\|(B_0+F)^{-1}\|\leq1/(\sqrt2-\rho)$ by the Neumann series. The resolvent identity gives \begin{equation} \|T-T_0\| \leq2\|(B_0+F)^{-1}\|\|F\|\|B_0^{-1}\| \leq\frac{\sqrt2\rho}{\sqrt2-\rho}. \label{cx:resolvent-proof} \end{equation} The triangle inequality gives $\|Ta\|\geq(1-\|T-T_0\|)\|a\|$. Solving the sufficient strict inequality $\rho/(1-\rho/\sqrt2)<1-\sqrt\tau$ gives \eqref{cx:rho-certificate}. \end{proof} \begin{remark}[Model generality and port errors] \label{cx:port-errors} The perturbation may mix different pairs and need not be diagonal. A sufficient unweighted condition is $\|\mathcal E\|/\min_jJ_j\to0$, since $\rho\leq\|\mathcal E\|/\min_jJ_j$; weighted control can be much sharper. If normalized input/output couplings also change to $E_L+L$ and $E_R+R$, and there is a direct transmission correction $D_0$, the same proof yields \begin{equation} \|T-T_0\|\leq \frac{\rho}{1-\rho/\sqrt2} +\frac{2(\|L\|+\|R\|+\|L\|\|R\|)}{\sqrt2-\rho} +\|D_0\|. \label{cx:all-errors} \end{equation} The normalization here is relative to the original $G$; for example the physical input matrix in the model is $\sqrt2G(E_L+L)$. An arbitrary algebraic perturbation is not automatically passive or reciprocal. Any proposed physical realization must also establish those properties and the power normalization of its actual ports. The estimate is a sufficient matrix certificate once those facts hold. \end{remark} \subsection{Compression and coherent passage certification} For a proposed device, the relevant test is a bound on every vector in the input subspace. The compressed transmission matrix and the section-flux matrix below give convenient sufficient tests for that purpose. The absolute-bypass test is stronger than a bound on net signed flux: return flow outside the section must be counted even where opposite contributions cancel. \begin{proposition}[Physical channel certificate if its data exist] \label{cx:compression} Let $T:\mathcal H_{\rm in}\to\mathcal H_{\rm out}$ be a bounded physical transmission operator between power Hilbert spaces. Let $V$ and $Z$ be isometries from $\mathbb C^N$ into its input and output spaces, and let $M=Z^*TV$. If $U$ is unitary and $\|M-U\|\leq e<1-\sqrt\tau$, then $\|TVa\|^2>\tau\|a\|^2$ for every nonzero $a$. If $T$ is a contraction, and its physical section-flux matrix on these inputs obeys $Q_D\succeq\eta I_N$, then the passage inequality $P_D(Va)\geq\eta\|TVa\|^2$ also holds for every $a$. \end{proposition} \begin{proof} The isometry $Z$ gives $\|TVa\|\geq\|Z^*TVa\|$; the unitary comparison gives $\|Ma\|\geq(1-e)\|a\|$. For the second statement, $P_D(Va)=a^*Q_Da\geq\eta\|a\|^2\geq\eta\|TVa\|^2$. The exact, potentially weaker passage test is $Q_D-\eta V^*T^*TV\succeq0$. None of these conclusions needs compactness or a discrete singular-value decomposition of the full operator. \end{proof} \begin{lemma}[A sufficient absolute-bypass certificate] \label{cx:absolute-bypass} Suppose actual basis fields have well-defined pointwise traces on $\Pi\setminus D$, where $\Pi$ is the section plane, and define the Hermitian matrix \begin{equation} F_{ij}(s)=\frac14\left(E_j(s)\times\overline{H_i(s)} +\overline{E_i(s)}\times H_j(s)\right)\cdot n. \label{cx:pointwise-flux} \end{equation} If these traces are regular enough that $\int_{\Pi\setminus D}\|F(s)\|\,dA$ is defined and at most $\delta$, then every coherent vector satisfies \begin{equation} \int_{\Pi\setminus D}|S_n(a;s)|\,dA\leq\delta\|a\|^2. \label{cx:bypass-certificate} \end{equation} \end{lemma} \begin{proof} Expansion gives $S_n(a;s)=a^*F(s)a$. Pointwise, $|a^*F(s)a|\leq\|F(s)\|\|a\|^2$; integrate. \end{proof} \begin{remark}[Limitations of basis-only and signed-flux tests] \label{cx:coherent-failure} Let $F(s)=h(s)\left(\begin{smallmatrix}0&L\\L&0\end{smallmatrix}\right)$, with $\int|h|=1$. Each coordinate basis vector has zero local flux, but $a=(1,1)^T/\sqrt2$ has absolute integrated flux $L$. If additionally $\int h=0$, the full \emph{signed} integrated flux matrix vanishes while the same absolute bypass remains $L$. This elementary example disproves both inference rules for Hermitian flux data; it is not asserted to be a global Maxwell realization. The evanescent pair in Proposition~\ref{cx:evanescent} gives an actual local Maxwell instance of individually zero-flux exponential fields acquiring power by interference. Likewise a basis test for signed passage can fail: $Q=\left(\begin{smallmatrix}1&2\\2&1\end{smallmatrix}\right)$ has positive diagonal but a negative eigenvalue. Section-flux conditions require a full matrix inequality; absolute bypass requires additional control such as Lemma~\ref{cx:absolute-bypass}. Rough Maxwell traces are not silently assumed to have these pointwise products. \end{remark} \subsection{Physical realization attempts and their specific obstructions} \begin{proposition}[A restricted radial angular-momentum obstruction] \label{cx:angular-obstruction} For a radial scalar dielectric with $\varepsilon(r)\leq\varepsilon_+$ on $0k_0^2\varepsilon_+R^2$ and the nonzero leading coefficient in $u\sim cr^{\ell+1}$ is positive, then $u,u'>0$ on $(0,R]$. In particular, neither $u(R)=0$ nor $u'(R)=0$ can hold. \end{proposition} \begin{proof} The regular solution also satisfies $u'(r)\sim(\ell+1)c r^\ell$. To justify this from the equation rather than differentiate an asymptotic without control, put $v=r^{-\ell-1}u$. Then $(r^{2\ell+2}v')'=-k_0^2\varepsilon(r)r^{2\ell+2}v$. The integration constant at zero must vanish for $v$ to have its finite limit $c$, so boundedness of $\varepsilon,v$ gives $v'=O(r)$. Thus both $u$ and $u'$ are positive for sufficiently small positive radius. As long as $u>0$, the displayed strict inequality gives $u''>0$, so $u'$ is increasing and remains positive. A first zero of either $u$ or $u'$ is consequently impossible. Across radial interfaces $u,u'$ are continuous; the same argument applies piecewise and hence globally. \end{proof} \noindent\textbf{Radial shells and partial waves.} Proposition~\ref{cx:bragg} proves that bounded-index reflectors can produce arbitrarily large internal amplitude. Repeating its exactly compensated shell profile for order $\ell$ would require $\varepsilon=n^2+\ell(\ell+1)/(k_0r)^2$. At a fixed inner radius this eventually violates the fixed material upper bound. Increasing the radius avoids that violation but does not place the high-order power through the original $D$. Proposition~\ref{cx:angular-obstruction} also blocks the particular real-core Neumann or Dirichlet tuning in a fixed small ball at high $\ell$. It is not a general exclusion of open resonances, nonspherical matching, or fields coupled from a larger exterior structure. The exact radial example is a one-port reflector; turning its stored energy into $N$ independently driven opposite-side ports is an unproved step. \medskip \noindent\textbf{Planar dielectric Bragg cavities.} At normal incidence, bounded-index quarter-wave layers do realize lossless one-dimensional mirrors; a symmetric resonant pair can realize the scalar feedback algebra. This statement follows by the same two-by-two transfer-matrix multiplication used above. Such an exact planar realization has infinite transverse extent and its plane waves are normalized per transverse area. It does not supply finite total incoming power with absolute flux confined to a prescribed finite $D$. Truncating the mirrors adds lateral leakage and diffraction. An $N$-cavity proposal must prove that these errors, along with off-diagonal mixing and detuning, satisfy the relative bounds of Theorem~\ref{cx:n-model} for the weakest direction. Merely increasing the number of mirror layers supplies no such multidimensional estimate. \medskip \noindent\textbf{Many small tips coupled to large remote resonators.} Arbitrarily fine dielectric tips or corners could create linearly independent local trace patterns. Linear independence does not bound the smallest singular value of the physical coupling matrix, and the field-to-output inverse is precisely where its weakness might be compensated. No bounded-index synthesis is established that makes all $N$ resonances coincide at the prescribed frequency while controlling leakage, parasitic coupling, and model error relative to that smallest coupling. Remote support growth is allowed, so its expense alone is not an obstruction. In contrast, assuming ideal independently tunable impedances or perfect conducting partitions would change the physical class. These routes remain candidates, not impossibility results. \medskip \noindent\textbf{Thin dielectric guides and broadened channels.} A weakly confined guide mode, when it exists, does not automatically satisfy the fixed-aperture bypass requirement. The prior report's qualified planar-guide construction explicitly changes the lower permittivity bound toward zero, varies the section shape with unbounded diameter, and uses infinite leads. It therefore cannot be transferred to the present fixed ordinary-dielectric, fixed-background, same-section problem. Counting outputs at different frequencies, adding internal sources, or sending the threshold $\tau$ to zero would likewise answer different questions. \begin{remark}[Open physical realization of the network model] \label{cx:missing-bridge} For each $N$ one would need an actual finite dielectric and power isometries $V_N,Z_N$ at the same frequency whose completed compression obeys the bound in \eqref{cx:all-errors} with a fixed margin below $1-\sqrt\tau$, together with the physical section inequality in Proposition~\ref{cx:compression} and an absolute-bypass estimate such as \eqref{cx:bypass-certificate}. In a resonator implementation this requires a justified Maxwell reduction and a weighted error matrix; $o(\min_jJ_j)$ unweighted error is sufficient but has not been proved. The matrices $J^{-1}$ and $I_N$ in Theorem~\ref{cx:n-model} are model energy and internal-current matrices, not computed Maxwell energy or section-flux matrices. If this bridge were established for arbitrary $N$, the fixed-parameter finite channel bound would be false. Its absence means that neither a disproof nor a universal bound has been obtained by this network-model track. The different physical construction in Theorem~\ref{uc:counterexample} gives a counterexample in the regime $\delta>\eta$ without establishing this model reduction. \end{remark}