Open questions and potential solutions

The extreme lifted phases move inward while their spread is below $\pi$, so the containing interval is invariant. Symmetry preserves the lifted mean. Pairing the two orientations of each edge gives a quadratic dissipation identity; the sine-sector inequality and the Laplacian Rayleigh bound yield $q^{\mathrm{\prime }}\le -2kq$ for the squared centered norm $q$. An integrating factor proves the exponential estimate. At short-arc entry, the variance bound $q\le n{\gamma }^2\mathrm{/}4$ makes the remaining time independent of the particular entering configuration. Applying this deterministic implication on the event $T_{\gamma }\le s$ proves the probability transfer. The numbered argument below also verifies an exact complete-graph family whose entrance times diverge near an antipodal two-cluster equilibrium.

Direct energy estimate and event inclusion. Establish interval invariance, conservation of the lifted mean, and a spectral dissipation inequality. Integrate that inequality, derive a uniform variance bound at cohesive entry, and transfer any entrance-time quantile to a synchronization-time quantile. Finally verify an exact delayed-entry trajectory to delimit what this partial theorem can establish. The argument is locally derived; the external consensus theorem supplies independent context rather than an unproved step.

1. The assigned question is attributed to Section 7 of the original article. Independent context is supplied by Dörfler–Bullo’s Theorem 4.8: specialize its positive coefficients $D_i$ to $1$, phase shifts to zero, equal frequency ratios to a common frequency, and its symmetric graph to a connected graph. Its exponential exponent becomes $-{\lambda }_2(L)\mathrm{sinc}(\gamma )$. These source statements motivate the calculation; neither the authors’ question nor their global-synchronization conclusions are premises of the proof below. [ref-speed-author-question, ref-speed-cohesive-consensus]
2. Choose the real rotating-frame lift at time $T$ and continue it using the stated differential equation. Write $M(t)={\max }_x{\varphi }_x(t)$ and $m(t)={\min }_x{\varphi }_x(t)$. On any interval where $M-m<\pi$, an active maximum has derivative $-{\sum }_y{A}_{x,y}\sin (M-{\varphi }_y)\le 0$, and an active minimum has derivative at least zero. The finite-envelope rule therefore makes $M$ nonincreasing and $m$ nondecreasing there. Starting with spread $\gamma <\pi$, continuity prevents a first exit to spread $\pi$. Thus $[m(T),M(T)]$ is invariant for all later times, and every lifted pair difference has absolute value at most $\gamma$. [hyp-speed-homogeneous-dynamics, lem-speed-finite-envelopes]
3. Symmetry and oddness of sine give ${\sum }_x{\dot{\varphi }}_x=0$, so $\bar{\varphi }$ is constant. Put $z=\varphi -\bar{\varphi }\mathbf{1}$ and $q={\sum }_x{z}_x^2$. Pairing ordered edges gives $q^{\mathrm{\prime }}=-{\sum }_{x,y}{A}_{x,y}({\varphi }_x-{\varphi }_y)\sin ({\varphi }_x-{\varphi }_y)$. For $0<\gamma <\pi$, the sine-sector inequality bounds this by $-\mathrm{sinc}(\gamma ){\sum }_{x,y}{A}_{x,y}(z_x-z_y{)}^2=-2\mathrm{sinc}(\gamma )z^{T}Lz$. Since $z\perp \mathbf{1}$, the Rayleigh bound gives $q^{\mathrm{\prime }}\le -2kq$. If $\gamma =0$, all coordinates already agree and their derivatives vanish. [hyp-speed-homogeneous-dynamics, def-speed-cohesive-error, lem-speed-sine-sector, lem-speed-laplacian-gap, step:2]
4. The integrating factor gives $\frac{d}{dt}(e^{2k(t-T)}q(t))\le 0$, hence $q(t)\le R^2{e}^{-2k(t-T)}$. Taking square roots and solving the exponential inequality proves the stated tail time. If $R=0$, the same inequality forces $q(t)=0$. The flat-torus distance to the synchronized set is at most $\mathrm{\parallel }z(t){\mathrm{\parallel }}_2$, by choosing the common phase represented by $\bar{\varphi }$, so the bound also guarantees permanent torus error at most $\epsilon$. [def-speed-cohesive-error, lem-speed-scalar-calculus, step:3]
5. For the expander specialization, assume the explicit matrix inequality ${\mathrm{\Delta }}_L=L-d^{\ast }I+(d^{\ast }\mathrm{/}n)J⪰c^{-}{d}^{\ast }I$, with $d^{\ast }>0$ and $c^{-}>-1$. For every centered real vector $f$, $Jf=0$, so $f^{T}Lf=d^{\ast }\mathrm{\parallel }f{\mathrm{\parallel }}_2^2+f^T{\mathrm{\Delta }}_{L}f\ge (1+c^{-})d^{\ast }\mathrm{\parallel }f{\mathrm{\parallel }}_2^2$. Minimizing over centered unit vectors yields ${\lambda }_2(L)\ge (1+c^{-})d^{\ast }>0$. Substitution gives the asserted sufficient tail bound without invoking any global-synchronization theorem. [hyp-speed-expander-lower-bound, lem-speed-laplacian-gap, step:4]
6. If lifted coordinates lie in $[a,b]$, let $\mu$ be their mean. Averaging $({\varphi }_x-a)(b-{\varphi }_x)\ge 0$ gives $n^{-1}{\sum }_x{\varphi }_x^2\le (a+b)\mu -ab$. Therefore $n^{-1}{\sum }_x({\varphi }_x-\mu {)}^2\le (\mu -a)(b-\mu )\le (b-a{)}^2\mathrm{/}4$. In particular, entry into an arc of length at most $\gamma$ permits a lift with centered norm at most $\sqrt{n}\gamma \mathrm{/}2$. [def-speed-cohesive-error, lem-speed-scalar-calculus]
7. The set $C_{\gamma }$ of phase configurations contained in some closed arc of length $\gamma$ is compact: it is the continuous image of $S^1\times [0,\gamma {]}^V$ under $(a,u)\mapsto (a+u_x{)}_x$. Thus a finite first entrance time is attained. On $\{T_{\gamma }\le s\}$, apply steps 4 and 6 at $T_{\gamma }$ to obtain ${\tau }_{\epsilon }\le T_{\gamma }+b\le s+b$. Consequently $\{{\tau }_{\epsilon }>s+b\}\subseteq \{T_{\gamma }>s\}$, proving the probability inequality for any initial-state law. In particular, an independently established estimate $\mathrm{Pr}(T_{\gamma }\le s)\ge 1-\delta$ would imply $\mathrm{Pr}({\tau }_{\epsilon }\le s+b)\ge 1-\delta$; no such entrance estimate has been assumed as an established fact. [def-speed-entrance-times, step:4, step:6]
8. To verify the entrance obstruction, take an even $n\ge 4$ and the loop-free complete graph $K_n$. Split its vertices into two equal groups with lifted phases $-\delta (t)\mathrm{/}2$ and $\delta (t)\mathrm{/}2$. Direct substitution shows that this is an exact solution when ${\delta }^{\mathrm{\prime }}=-n\sin \delta$. With $\delta (0)=\pi -\eta$ and $0<\eta <\pi -\gamma$, integration gives $\tan (\delta (t)\mathrm{/}2)=\cot (\eta \mathrm{/}2)e^{-nt}$ and $T_{\gamma }=n^{-1}\log (\cot (\eta \mathrm{/}2)\mathrm{/}\tan (\gamma \mathrm{/}2))$. Before entry, the minimal containing arc has length $\delta (t)$ because $0<\delta (t)<\pi$. Hence $T_{\gamma }\to \mathrm{\infty }$ as $\eta \downarrow 0$. Meanwhile $A=J-I$, $d^{\ast }=n-1$, ${\mathrm{\Delta }}_A=J\mathrm{/}n-I$, and $\mathrm{\parallel }{\mathrm{\Delta }}_A\mathrm{\parallel }=1$, so $K_n$ has adjacency-expansion parameter $\alpha =1\mathrm{/}(n-1)$; also $L=nI-J$ and ${\lambda }_2(L)=n$. Thus arbitrarily strong expansion across this family does not bound entrance time uniformly over all initial phases. These specially chosen initial states do not determine the entrance-time distribution for uniform random phases. [hyp-speed-homogeneous-dynamics, hyp-speed-expander-lower-bound, def-speed-entrance-times, lem-speed-scalar-calculus]

• ref-speed-author-question: Attribution only: the authors ask how quickly a random initial state synchronizes on sufficiently expanding graphs. Their homogeneous model uses symmetric adjacency and a common intrinsic frequency. No speed estimate or global-synchronization theorem is imported here. Source: https://static1.squarespace.com/static/5436e695e4b07f1e91b30155/t/6991d638ab1b273169a7becd/1771165240824/expander-graphs-are-globally-synchronizing.pdf; PDF p. 39, Section 7, penultimate paragraph; pp. 2–3, equation (1.1) and Definition 1.1.. Reading limits: Freshly read these extracted PDF passages and inspected the retained p. 39 image. This direction does not claim a fresh reading of all article proofs or verification of its global-synchronization results.
• ref-speed-cohesive-consensus: Independent context: Theorem 4.8 assumes a globally reachable node, zero phase shifts, equal frequency ratios and initial phases in an arc of length less than pi. For symmetric coupling it gives convergence to the coefficient-weighted mean with the spectral exponent in equation (4.23). Setting all coefficients $D_i$ to one gives exponent minus lambda_2(L) sinc(gamma). Source: https://motion.me.ucsb.edu/pdf/2009z-db.pdf; PDF p. 22, Section 4.3, Theorem 4.8, equation (4.23), and the immediately following proof.. Reading limits: Freshly read the primary PDF theorem and its proof passage through the time-varying consensus reformulation. The proof refers back to Theorem 4.1; that earlier argument was not reconstructed. The local proof here independently supplies the symmetric homogeneous specialization.
• hyp-speed-homogeneous-dynamics: Given a finite undirected graph, A is symmetric and entrywise nonnegative, D has diagonal entries sum_y $A_{x}y$, and $L=D$-A. A global C1 real rotating-frame trajectory satisfies phi'_x=-sum_y $A_{x}y$ sin(phi_x-phi_y). At the specified time T a chosen lift has diameter gamma<pi. Loops contribute zero to both phase differences and the Laplacian quadratic form. Source: self-contained; Solution statement; proof steps 2–3 and 8.. Reading limits: These are explicit mathematical hypotheses. Existence of the given trajectory is assumed; no physical observation, open problem or conclusion of the article supplies it.
• def-speed-cohesive-error: For a real lift, the centered error is z=phi-($n^{-1}$sum_x phi_x)1 and its Euclidean norm is the lifted error. On the phase torus define $d_S$(theta)=inf_c (sum_x min_{j in Z}|theta_x-c+2pi j|^2)^{1/2}. This distance to the synchronized set is no larger than the centered error of any real lift. Source: self-contained; Solution statement; proof steps 3–4 and 6.. Reading limits: Definitions and the comparison follow directly by choosing c equal to the lifted mean. No published supplier or independent novelty claim is attached to them.
• lem-speed-finite-envelopes: The maximum and minimum of finitely many C1 functions are locally Lipschitz. Their right directional derivatives are respectively the maximum and minimum derivatives among active coordinates. This follows by first-order expansion and exclusion of inactive coordinates with a strict value gap. A locally Lipschitz function with derivative nonpositive almost everywhere is nonincreasing, by absolute continuity. Source: self-contained; Proof step 2.. Reading limits: The finite-envelope argument is supplied here. The elementary absolute-continuity integration rule is used; no external consensus theorem is needed for interval invariance.
• lem-speed-sine-sector: For 0<gamma<pi and |u|$\le$gamma, u sin(u)$\ge$sinc(gamma)$u^2$. Indeed sinc is even, and on (0,pi) its derivative has numerator u cos(u)-sin(u)<0: sin(u)-u cos(u) has derivative u sin(u)>0 and value zero at zero. The value at zero is obtained by continuity. Source: self-contained; Proof step 3.. Reading limits: The scalar inequality is locally derived from elementary trigonometric differentiation, rather than imported from the article or consensus paper.
• lem-speed-laplacian-gap: For symmetric nonnegative A, $f^{T}Lf$=(1/2)sum_xy $A_{x}y$($f_x$-$f_y$)^2. For a connected graph on at least two vertices, lambda_2(L)=min{$f^{T}Lf$: sum_x $f_x$=0, ||f||_2=1}>0. The identity follows by expansion; zero form forces constancy along every edge and hence throughout the graph. Compactness of the centered unit sphere then makes its minimum positive. This Rayleigh value equals the second-smallest Laplacian eigenvalue by diagonalization of a real symmetric matrix. Source: self-contained; Proof steps 3, 5 and 8.. Reading limits: The quadratic identity and positivity argument are supplied here. The conventional eigenvalue interpretation uses the elementary finite-dimensional real symmetric spectral theorem; it is not an assumption about an expander’s global dynamics.
• lem-speed-scalar-calculus: The elementary analysis inputs are the product and chain rules, integration of a differential inequality by an exponential integrating factor, tan(delta/2)'/tan(delta/2)=delta'/sin(delta) for 0<delta<pi, and ab$\le$((a+b)/2)^2 for nonnegative a,b. Finite-dimensional closed bounded sets are compact. Source: self-contained; Proof steps 4 and 6–8.. Reading limits: These elementary identities and compactness facts are the analytic inputs to the displayed local calculations. No numerical experiment is used as proof.
• hyp-speed-expander-lower-bound: For the optional expander specialization assume d^>0, $c^{-}$>-1 and Delta_L=L-d^I+(d^/n)J $\ge$ $c^{-}d$^I in quadratic-form order. Here $J=11$^T. For comparison with adjacency expansion, Delta_A=A-(d^/n)J and ||Delta_A||$\le$alpha d^. These are explicit definitions and hypotheses, not consequences of a synchrony claim. Source: self-contained; Proof steps 5 and 8; notation agrees with the original article’s PDF p. 6, Section 1.2, and p. 8, Definition 1.9.. Reading limits: Freshly checked the cited original definitions and the p. 8 image for notation. The needed spectral implication is proved directly in step 5, without using the original article as an independent proof supplier.
• def-speed-entrance-times: For fixed 0<gamma<pi, $C_{g}amma$ consists of configurations contained in a closed circular arc of length gamma. Define $T_{g}amma$=inf{t$\ge$0: theta(t) belongs to $C_{g}amma$}, with infimum infinity when entry never occurs. Define tau_epsilon=inf{t$\ge$0: $d_S$(theta(u))$\le$epsilon for every u$\ge$t}. Random initial phases use the product uniform measure when addressing the authors’ question; the event-inclusion argument applies to any initial law. Source: self-contained; Solution statement; proof steps 7–8.. Reading limits: These are local definitions. Compactness gives attainment of finite entry. Measurability follows from continuity of the finite-dimensional flow and countable dense-time tests; no quantitative entrance-time law is asserted.

Decision: accept

Independently checked all eight argument steps and ten immediate dependencies. Interval invariance, mean conservation, spectral dissipation, the variance estimate and event inclusion establish the stated conditional theorem. The complete-graph calculation correctly demonstrates unbounded deterministic entrance delays without implying a lower bound for uniformly random initialization. The partial/proved classification is appropriate.

Dependency review: The mathematical dependencies are explicit and non-circular. I checked the finite-envelope rule, sine-sector inequality, Laplacian quadratic form and positive spectral gap, integrating-factor calculation, variance bound, compactness/attainment argument and measurability through continuous-flow dense-time tests. The centered-Laplacian hypothesis directly yields lambda_2(L) $\ge$ (1+$c^{-}$)d^*. For the complete-graph example, substitution gives delta'=-n sin(delta), the stated entrance time, ||Delta_A||=1 and lambda_2(L)=n. The source specialization agrees with Dörfler–Bullo, Theorem 4.8 and equation (4.23): setting their coefficients $D_i$=1 gives the asserted exponent. That theorem provides context; the local argument independently proves the needed specialization. The original article supplies question attribution and notation, not proof of its own prerequisites. No physical claim or open problem is used as a mathematical premise.

Review reading limits: Freshly read the frozen direction, every numbered step and every immediate dependency; original PDF text on pp. 2–3, 6, 8 and the question paragraph on p. 39; retained original images on pp. 8 and 39; and Dörfler–Bullo’s PDF p. 22, Theorem 4.8, equation (4.23) and its proof paragraph. I did not reconstruct its referenced Theorem 4.1. Verified the supplied verification artifact’s SHA256 and inspected its reading limits; its broader checks were not adopted as my own. This is the assigned direction review, not a whole-article defect assessment or exhaustive novelty search.

The original solution uses defective expanders and phase-set amplification, followed by a coupled process argument covering near-connectivity and later times. The independent derivation below instead combines the uniform normalized-spectrum theorem of Hoffman–Kahle–Paquette with McRae’s normalized condition-number criterion. An isolated-vertex estimate places connectivity after the spectral theorem’s starting time; the spectral window then gives condition number below two simultaneously thereafter.

Preserve the established solution’s attribution, then give a separate direct composition. Couple edge indices to exponential clocks, show connectivity occurs after the deterministic spectral starting time, intersect two high-probability events, and apply a deterministic synchronization criterion uniformly along each resulting path.

1. The published solution matches the requested simultaneous quantifier. Its historical method uses defective expanders and amplification, followed by a two-regime process application. This attribution is separate from the derivation in steps 2–8. [input-jain-mizgerd-sawhney]
2. Assume countable choice. Assign independent uniform labels $U_e\in [0,1)$ to the possible edges and set $T_e=-\log (1-U_e)$. These are independent rate-one exponential clocks. Writing $H_t=G_{p(t)}$ with $p(t)=1-e^{-t}$, their almost surely strict edge ordering has the prescribed process law. If $s_m$ is the $m$th arrival time, then $H_{s_m}=G(n,m)$ pathwise; $s_{\tau }$ is the first connected clock time. [math:def-countable-choice, def-expander-prereq-edge-order-process-coupling, lem-author-fc0d6c95270f-exponential-process-transfer]
3. Put $p_0=3\log n\mathrm{/}(4n)$ and $t_0=-\log (1-p_0)$ for sufficiently large $n$. To see why connectivity occurs later, let $X$ count isolated vertices in $G_{p_0}$ and set $q=(1-p_0{)}^{n-1}$. Independence gives $\mathbb{E}X=nq$ and $\mathrm{Var}X\mathrm{/}(\mathbb{E}X{)}^2\le 1\mathrm{/}(nq)+p_0\mathrm{/}(1-p_0)$. The supplied local argument proves $nq\ge n^{1\mathrm{/}8}$ eventually. Since $(X-\mathbb{E}X{)}^2=(\mathbb{E}X{)}^2$ on $\{X=0\}$, this bounds $\mathbb{P}(X=0)$ by a quantity tending to zero. Thus an event $A_n$ of probability tending to one satisfies $s_{\tau }>t_0$. [lem-author-fc0d6c95270f-isolated-vertex-before-onset, step:2]
4. Apply the imported uniform spectral theorem with $k=1$ and $\delta =1\mathrm{/}4$. It supplies a constant $C$, independent of $n$, and events $E_n$ of probability tending to one that control every $t\ge t_0$ together. At connected times no isolated vertices are removed, so the resulting bound applies to the full normalized Laplacian of $H_t$. [input-hkp-uniform-spectrum, def-author-fc0d6c95270f-normalized-laplacian, step:2]
5. On $A_n\cap E_n$, every $t\ge s_{\tau }$ is connected and lies after $t_0$. Hence $\eta (H_t)<r_n:=C\mathrm{/}\sqrt{(n-1)p_0}$ simultaneously for all such times. Because $(n-1)p_0\to \mathrm{\infty }$, eventually $r_n<1\mathrm{/}3$. Also $\mathbb{P}(A_n\cap E_n)\ge 1-\mathbb{P}(A_n^c)-\mathbb{P}(E_n^c)\to 1$. This uses two pathwise events, without evaluating a fixed-time probability estimate at a random index. [def-expander-prereq-edge-order-process-coupling, step:3, step:4]
6. For every graph on this path segment, $1-r_n\le {\lambda }_2(\mathcal{L})\le {\lambda }_n(\mathcal{L})\le 1+r_n$, so ${\lambda }_n(\mathcal{L})\mathrm{/}{\lambda }_2(\mathcal{L})\le (1+r_n)\mathrm{/}(1-r_n)<2$. Connectedness and $n\ge 2$ give positive degrees. Moreover $x^{T}Lx={\sum }_{\{i,j\}\in E}(x_i-x_j{)}^2$ vanishes exactly for constant vectors, so the symmetric unnormalized Laplacian has ${\lambda }_2(L)>0$. Thus all hypotheses of the deterministic criterion hold. [def-author-fc0d6c95270f-normalized-laplacian, lem-author-fc0d6c95270f-spectral-window-ratio, step:5]
7. The imported deterministic criterion now makes every $H_t$ with $t\ge s_{\tau }$ globally synchronizing on the same event. Its coupling matrix is the simple graph adjacency matrix and its positive global coupling constant is taken to be one. Substituting each arrival time $s_m$ transfers the conclusion to every $\tau \le m\le N$ simultaneously, establishing the stated probability limit. [input-mcrae-normalized-criterion, lem-author-fc0d6c95270f-exponential-process-transfer, step:2, step:5, step:6]
8. The outer probability concerns graph paths. For a fixed successful path, each of its finitely many post-connectivity graphs has a phase-null exceptional set. Their finite union remains phase-null. Consequently one may even choose a common full-volume set of initial configurations that synchronize when the autonomous flow is separately restarted on each graph of that path. This finite-intersection observation makes no assertion about switching the graph during a trajectory. [def-expander-prereq-phase-volume, step:7]

• math:def-countable-choice: Assume that every sequence of nonempty sets admits a choice function, as required by the local probability and phase-measure suppliers. Source: math:def-countable-choice; Definition. Reading limits: Read the supplier’s metadata, Definition and Remarks. Its transitive suppliers and cited literature were not independently reviewed.
• def-expander-prereq-edge-order-process-coupling: The process is a uniform ordering of the possible edges; independent continuous uniform labels realize that ordering. Connectivity persists under edge addition. High probability means probability tending to one, and the post-connectivity assertion is a single simultaneous event. Source: def-expander-prereq-edge-order-process-coupling; Definition; Justification, steps 1.1–3.1. Reading limits: Read the frozen formulation and local coupling argument. No independent review of its transitive library closure is claimed.
• def-expander-prereq-phase-volume: Normalized phase volume is normalized Lebesgue measure on a fundamental angular box pushed to the phase torus. Almost everywhere excludes a measurable null set; finite unions of such exceptional sets remain null. Source: def-expander-prereq-phase-volume; Definition. Reading limits: Read the frozen definition and metadata. The measure construction and transitive suppliers were not re-proved.
• def-author-fc0d6c95270f-normalized-laplacian: For a simple graph with positive degree matrix $D$, define $L=D-A$ and $\mathcal{L}=D^{-1\mathrm{/}2}L{D}^{-1\mathrm{/}2}$. For a connected graph, define $\eta (G)={\max }_{i>1}\mathrm{\mid }1-{\lambda }_i(\mathcal{L})\mathrm{\mid }$ and $\kappa (G)={\lambda }_n(\mathcal{L})\mathrm{/}{\lambda }_2(\mathcal{L})$. Source: def-author-fc0d6c95270f-normalized-laplacian; Definition; Justification, steps 1.1–2.1. Reading limits: Read the frozen definition and its local kernel and spectral justification; no transitive supplier audit is claimed.
• lem-author-fc0d6c95270f-exponential-process-transfer: Under countable choice, $T_e=-\log (1-U_e)$ gives independent rate-one exponential clocks, preserves the uniform edge ordering, and identifies each indexed graph with the graph at its arrival time. A pathwise all-time assertion after connectivity transfers to every later edge index. Source: lem-author-fc0d6c95270f-exponential-process-transfer; Statement; Proof, steps 1.1–4.1. Reading limits: Read the complete frozen statement and local proof.
• lem-author-fc0d6c95270f-isolated-vertex-before-onset: Under countable choice, at $p_0=3\log n\mathrm{/}(4n)$ the independent-label graph has an isolated vertex with probability tending to one. Consequently its first connected clock time exceeds $t_0=-\log (1-p_0)$. The proof gives $\mathbb{P}(X=0)\le n^{-1\mathrm{/}8}+p_0\mathrm{/}(1-p_0)$ eventually. Source: lem-author-fc0d6c95270f-isolated-vertex-before-onset; Statement; Proof, steps 1.1–5.1. Reading limits: Read and checked the finite second-moment calculation and the elementary lower bound for the expected isolated-vertex count.
• lem-author-fc0d6c95270f-spectral-window-ratio: For a connected graph, a normalized spectral window $\mathrm{\mid }1-{\lambda }_i\mathrm{\mid }\le r<1\mathrm{/}3$ for every $i>1$ implies $\kappa (G)\le (1+r)\mathrm{/}(1-r)<2$. Source: lem-author-fc0d6c95270f-spectral-window-ratio; Statement; Proof, steps 1.1–2.1. Reading limits: Read the complete frozen argument and checked its inequalities.
• input-jain-mizgerd-sawhney: For the empty-start uniform missing-edge process, with high probability all graphs at indices $m\ge \tau$ are simultaneously globally synchronizing. This source supplies the historical attribution and original proof outline, rather than a premise of the independent composition. Source: https://arxiv.org/html/2501.12205v1; Section 1, process definition, preceding quantifier remark and Theorem 1.2; Theorem 1.6; Sections 2.1, 2.3–2.4 and 3. Reading limits: Freshly read these HTML passages, including the final process proof. The underlying concentration estimates and amplification proof were not independently reconstructed or audited.
• input-hkp-uniform-spectrum: For the independent rate-one exponential edge process, let $p(t)=1-e^{-t}$ and $d(t)=(n-1)p(t)$. Specialize Theorem 6.1 to $k=1$ and fixed $\delta =1\mathrm{/}4$, with $p(t_0)=3\log n\mathrm{/}(4n)$. A constant $C$, independent of $n$, gives, with high probability, ${\max }_{i>1}\mathrm{\mid }1-{\lambda }_i\mathrm{\mid }<C\mathrm{/}\sqrt{d(t)}$ simultaneously for every $t\ge t_0$, after deleting isolated vertices. Source: https://arxiv.org/pdf/1201.0425v6; Section 6, PDF p. 22, process, graph specialization and starting-time definitions; PDF p. 23, Theorem 6.1. Reading limits: Read PDF pp. 22–23 as text and inspected the rendered theorem on p. 23. Also read Theorem 1.3 and its following discussion. Imported the exact theorem; its proof and supporting estimates were not reconstructed.
• input-mcrae-normalized-criterion: Let $A$ be symmetric with $A\mathbf{1}$ strictly positive, $D=\mathrm{diag}(A\mathbf{1})$, $L=D-A$ and $\mathcal{L}=I-D^{-1\mathrm{/}2}A{D}^{-1\mathrm{/}2}$. If ${\lambda }_2(L)>0$ and ${\lambda }_n(\mathcal{L})\mathrm{/}{\lambda }_2(\mathcal{L})<2$, the homogeneous Kuramoto network globally synchronizes. Its global coupling constant may be any fixed $K>0$; here $K=1$. Source: https://arxiv.org/pdf/2503.18801v2; PDF p. 3, Section 1.2, equation (2) and convergence convention; PDF p. 10, Theorem 3.4, positive implication; PDF p. 11, Section 3.4.2. Reading limits: Read these PDF text passages and inspected the rendered Theorem 3.4 on p. 10 to distinguish ordinary and normalized Laplacians. Did not reconstruct Theorem 2.1, Section 6.1 or the cited analytic dynamical-systems results.

Decision: accept

Checked all eight argument steps. The exponential-clock coupling preserves the uniform edge process. The isolated-vertex calculation places connectivity after the deterministic spectral starting time with high probability. The imported all-time spectral event then gives a normalized condition number below two simultaneously after connectivity. Connectedness supplies positive degrees and a positive ordinary Laplacian gap, meeting the deterministic synchronization criterion. The finite union of phase-null exceptional sets is valid. The conclusion concerns separately run autonomous flows on fixed graphs. The original article’s p. 38 attribution to Jain, Mizgerd and Sawhney is accurate; their defective-expander method remains distinct from the later normalized-Laplacian derivation.

Dependency review: Checked every listed immediate dependency for its use here. The local clock, second-moment, kernel and spectral-ratio arguments are applicable and non-circular. The declared dependency closure contains no cycle. These are mathematical premises; no physical observation or open problem supplies a proof premise. Hoffman–Kahle–Paquette, Theorem 6.1, specialized to $k=1$ and $delta=1$/4, explicitly controls every later time on one high-probability event. Removing isolated vertices makes no change after connectivity. McRae, Theorem 3.4 permits the normalized ratio criterion with the separately verified ordinary Laplacian gap. Jain–Mizgerd–Sawhney, Theorem 1.2 states the simultaneous conclusion and supplies historical attribution rather than a premise of the independent composition.

Review reading limits: Freshly read the frozen direction and its immediate local supplier bodies, plus the two supporting Laplacian propositions and countable-choice definition. Inspected the original article’s p. 38 image. Read Hoffman–Kahle–Paquette’s Section 6 definitions and Theorem 6.1, including its rendered theorem; read McRae’s dynamics convention, Theorem 3.4 and Section 3.4.2, including the rendered criterion. Read Jain–Mizgerd–Sawhney’s theorem, strategy and process-proof passages. Imported theorem proofs and their supporting estimates were not independently reconstructed. No transitive library proof audit is claimed. Consulted the verification artifact’s reading limits and relevant hitting-time inventory records, without adopting its broader review coverage. The draft’s inherited full-article source-reading description does not describe this reviewer’s reading. This decision covers only the assigned direction.