\section{Uniform thresholds of order \texorpdfstring{$6p$}{6p}}\label{sec:six-p} We retain the definitions of $K_t(B)$, full-period height and digit length from the preceding sections. This family shares the three-row geometry of $3p$, but opposite roots change both the operator and the ordered capacity boxes. The critical-prefix proof below is scalar and deductive; its complete boundary calculations are included in Appendix~\ref{app:6p-scalar}. Write $B=4m+r$, $0\le r<4$. For primes $p\ge5$, set \[ d_p=\begin{cases}2,&p\equiv5\pmod6,\\4,&p\equiv1\pmod6,\end{cases} \qquad e_p=\begin{cases}2,&p=5,7,\\4,&p\ge11.\end{cases} \] Define the four offsets \begin{equation}\label{eq:6p-offsets} (c_{p,0},c_{p,1},c_{p,2},c_{p,3}) =(3p+d_p,\ 5p-e_p,\ 6p+1,\ 8p+1). \end{equation} \begin{theorem}[Uniform order-$6p$ threshold]\label{thm:six-p} For every prime $p\ge5$, \[ \boxed{K_{6p}(4m+r)=6pm+c_{p,r}} \] whenever the base satisfies the corresponding bound: \[ \begin{array}{c|rrrr} r&0&1&2&3\\\hline B\text{ at least}&16&13&10&19. \end{array} \] In particular all four formulas hold for every $B\ge16$. The minimal eventual period of the coefficient functions is 4. No minimal base cutoff is asserted. \end{theorem} Relative to the common slope, the same formula is \[ K_{6p}(B)=\frac{3p}{2}B+ \begin{cases} 3p+d_p,&B\equiv0\pmod4,\\ 7p/2-e_p,&B\equiv1\pmod4,\\ 3p+1,&B\equiv2\pmod4,\\ 7p/2+1,&B\equiv3\pmod4. \end{cases} \] The specialization $p=5$ gives the order-30 formula in Corollary~\ref{thm:15-30}; the theorem treats the entire prime family. \begin{corollary}[Base ten] For every prime $p\ge5$, $K_{6p}(10)=18p+1$. Thus the lengths for orders $30,42,66,78$ are respectively $91,127,199,235$. \end{corollary} \subsection{Opposite roots, signed boxes and short elements} Put $n=3p$, $t=2n$, take a primitive $t$-th root $\xi$, and put $\zeta=-\xi$, of order $n$. Let $A=\Z[\xi]=\Z[\zeta]$. With \[ \mathcal H_B=(B-\xi)(B-\xi^{-1}) =B^2+1+B(\zeta+\zeta^{-1}), \] the lattice of the double-evaluation kernel is $\Lambda=\mathcal H_BA$. This follows from Propositions~\ref{prop:image}, \ref{prop:ideals}, and \ref{prop:delta}: here $\Phi_{6p}(1)=\Phi_{6p}(-1)=1$, so $\delta=1$. The reversal's power of $B$ does not change the kernel. Choose $\eta_3,\eta_p$ with $\zeta=\eta_3\eta_p$. Explicit choices are $\eta_3=\zeta^{p(p^{-1}\bmod3)}$ and $\eta_p=\zeta^{3(3^{-1}\bmod p)}$. The root at digit position $s$ is \[ \xi^s=(-1)^s\eta_3^{s\bmod3}\eta_p^{s\bmod p}. \] Represent elements of $A$ by full $3\times p$ arrays. Equal algebraic values differ by a row potential plus a column potential; in the anchored basis their coordinates are $F_{ij}-F_{i,p-1}-F_{2,j}+F_{2,p-1}$. At length $k=tL+a$, each residue modulo $t$ has capacity $(B-1)(L+\one_{sq_-$, their sum is at least $h+2$. This is impossible: the four-marker sum is at most $h+1$. For $r=0$ this follows from $a\le3p+d_p<4p$ and Lemma~\ref{lem:6p-gaps}; for $r=1$ it is the trivial bound four; for $r=2$ at most one marker is present; for $r=3$, $a\le2p+1<3p$ and the lemma gives at most two. Similarly, if $q_->q_+$, every transition $1\to0$ is impossible, using $a_j^-+b_{j+1}^+$. A nonconstant cyclic word needs both transition types. Thus the parameters coincide, with the stated ranges. \end{proof} On the branch $r=1$, the common parameter is forced to be 2. \subsection{Scalar rules and their six-step bulk patterns}\label{sec:6p-rules} We use six relations between consecutive bits $x,y$: \[ \begin{array}{c|cccccc} \text{symbol}&\EE&\RelA&\ZZ&\OO&\UU&\DD\\\hline \text{meaning}&x=y&\text{any }x,y&x=y=0&x=y=1&x\le y&x\ge y. \end{array} \] These are ordinary scalar constraints, not matrices or software states. For the common parameter $q$, the allowed pair $x=f_j,y=f_{j+1}$ is given by four tests: \begin{equation}\label{eq:6p-four-tests} \begin{array}{c|l} 00&b_j^-\ge h-q,\quad b_{j+1}^+\ge h-q\\ 01&b_j^-\ge h-q+1,\quad a_{j+1}^+\ge q\\ 10&a_j^-\ge q,\quad b_{j+1}^+\ge h-q+1\\ 11&a_j^-\ge q-1,\quad a_{j+1}^+\ge q-1. \end{array} \end{equation} All tables below are obtained by these four substitutions. Their entries involve only equality, a fixed bit, or a one-way inequality, so their consequences can be read directly. For completeness, the marker rule requires no CRT search. In column $j$, visit number $k$, $0\le k<6$, marks row $(j+kp)\bmod3$, positively if $j+k$ is even and negatively if it is odd. A column with $b$ visits uses $k=0,\ldots,b-1$. At the two boundaries, with $\varepsilon=0$ for upper and 1 for lower, the visit counts are \begin{equation}\label{eq:6p-visits} \begin{array}{c|l} r=0&3+\one_{j0$. The signed intervals $[-h_0,h_0]$ admit an integer lift and hence folded coefficients in $[0,h_0]^{6p}$. For the lower bound suppose a false positive has folded height $u\le h_0-1$. For $B\ge11$, \[ \frac{u}{(B-1)^2}\le\frac{B^2+B}{4(B-1)^2}\le\frac13, \] the last comparison being $B^2-11B+4\ge0$. For $B=10$, $u\le27$ and $27/81=1/3$. The strict inverse bound puts $\beta$ in the short shell in either case. Every nonconstant cyclic binary sequence has a selected position with unselected predecessor and an unselected position with selected predecessor. Choosing those two columns gives a critical cycle with $v=1$. Its capacity is $4u$, so $4u\ge B^2+B+1$, contradicting $u\le h_0-1$. \end{proof} For independent coefficient height $H\ge1$, the earlier complete-period bridge immediately yields \[ 6p\left(\ceil{h_{6p}(B)/H}-1\right)+1 \le K_{6p,H}(B)\le6p\ceil{h_{6p}(B)/H}. \] This is a bound for arbitrary $H$, not an exact formula for every partial period. \subsection{Exponential multiplicity of the first algebraic values}\label{sec:6p-multiple} Here we count distinct normalized values $\beta=\mathcal H_B^{-1}P(\xi)$ at the first threshold, not digit words or orbits under rotations. \begin{theorem}\label{thm:6p-multiplicity} Let $B\equiv1\pmod4$, $B\ge13$. For $p=5,7$ there is one such value. For primes $p\ge11$ the number is \[ \boxed{2^{\lambda_p}-1},\qquad \lambda_p=\begin{cases}(p-8)/3,&p\equiv5\pmod6,\\(p-10)/3,&p\equiv1\pmod6.\end{cases} \] More explicitly: \begin{itemize} \item If $p\equiv5\pmod6$, $i=1$ and $S$ is a nonempty union of the disjoint triples $\{2+3j,3+3j,4+3j\}$, $0\le j<\lambda_p$. \item If $p\equiv1\pmod6$, $i=0$ and the complement of $S$ is a nonempty union of $\{3+3j,4+3j,5+3j\}$, $0\le j<\lambda_p$. \end{itemize} \end{theorem} \begin{proof} At the upper length on this branch the maximum capacity is the same as at the preceding length. Thus the strict shell reduction remains exhaustive for every minimum word. Lemma~\ref{lem:6p-prefix} and Appendix~\ref{app:6p-r1} classify every admissible support: the free triples are independent, and the constant choice is excluded. For $p=5,7$, the appendix gives the single exceptional support. Each nonconstant support satisfies all critical constraints, and the noncritical bounds in Theorem~\ref{thm:six-p} supply its integral lift. The short-shell parametrization is injective, so no normalized values are counted twice. \end{proof} For example the numbers at $p=11,13,17,19,23,29$ are $1,1,7,7,31,127$. The exponential number of values is compatible with constructing one value directly from a single triple. This does not enumerate all minimum digit words. \begin{corollary}[Uniqueness on the other three branches]\label{cor:6p-unique} For every prime $p\ge5$, there is exactly one normalized algebraic value $\beta$ at the first threshold on each of the following domains: \[ \begin{array}{c|ccc} B\bmod4&0&2&3\\\hline B\text{ at least}&16&22&19. \end{array} \] Together with Theorem~\ref{thm:6p-multiplicity}, this classifies the number of normalized minimum values in the four stated ranges. It does not classify the minimum digit words. \end{corollary} \begin{proof} Lemma~\ref{lem:6p-prefix} gives precisely one nonconstant support on each of these three branches. Its scalar classification is completed in Appendices~\ref{app:6p-r0}, \ref{app:6p-r2} and \ref{app:6p-r3}. It remains to make the short shell exhaustive \emph{at} $K$, rather than only at $K-1$. The maximal capacity divided by $(B-1)^2$ is respectively \[ \frac{B+4}{4(B-1)},\qquad \frac{B+6}{4(B-1)},\qquad \frac{B+5}{4(B-1)} \] on residues $0,2,3$. These ratios are at most $1/3$ in the stated domains. The strict inverse bound places every nonzero normalized value in the short shell. Its parametrization is injective, and the upper construction supplies a lift of the single support. \end{proof} For $B=10,14,18$, the order-$6p$ threshold remains covered by Theorem~\ref{thm:six-p}, but the preceding corollary makes no uniqueness claim. The capacity increase at $K$ prevents reuse of the $K-1$ shell bound without further work. \subsection{Computational cross-checks} The companion retains the independent matrix reconstruction (F6) and integer path-count identities (F7) as regression checks on the prefix and multiplicity formulas. They are not premises of the scalar proof above. Component F8 checks the scalar boundary classification directly against selected/unselected inequalities on a finite panel, and checks the digit-transfer and reversal consequences below. All such panels are finite diagnostics; the uniform conclusions use the written gap, boundary and cyclotomic arguments. The certificate dependencies for orders $105,210$ are separate and remain as stated in Section~\ref{sec:repro}.