1. Theorem 7, the only non-trivial theoretical result, is not proven; Proposition 6 is both trivially true and incorrectly argued.
(a) Rounding term. The proof (A.5) asserts that the integer-optimal allocation satisfies ∣Bk⋆−DkL∣≤1 "by concavity of gk⋆(Bk)". This is unjustified twice over. First, even if the value were separable across contexts (which it is not, see (b)), the upper level would be an integer program with a single weighted constraint ∑kfkBk≤Btot, i.e. a knapsack; for non-uniform fk the integer optimum need not be within one unit of the continuous optimum in every coordinate. Second, the concavity of the hard-constraint value gk⋆(Bk) (Lemma 9) is "proved" by the statement that "the (Bk+2)-th best arm is worse than the (Bk+1)-th best", which is a myopic argument with no dynamic content; dropping one activation changes the trajectory of that arm and hence all subsequent decisions, so there is no coupling that keeps the two systems within one arm of each other. The same flaw underlies the claim that each misallocated unit costs at most Rmax per period in the long-run average.
(b) Index term. The proof says "given B⋆, each context k faces a standard restless bandit with N arms and budget Bk⋆" and invokes the exponential bound of Gast et al. per context. But the CBB is a single Markov system whose per-period kernel is drawn i.i.d.; arm states carry over across contexts, so there is no per-context RMAB to which a time-homogeneous asymptotic-optimality theorem applies. Section 7 itself states that no prior index theory addresses this setting. Moreover, Appendix B "verifies" UGAP and GES by analysing a scalar linear ODE with the LP-optimal actions at⋆,at⋆⋆ held fixed (Eqs. 25–27). Under a priority policy the actions applied to a type-state category depend on how many arms occupy higher-priority categories, so the mean-field dynamics are piecewise-linear and coupled across types through the budget; the constant-action linearisation is precisely the local object that UGAP is not equivalent to. Nothing in Appendix B establishes the hypotheses of the cited theorem.
(c) Units. The left-hand side of Eq. (4) is the total average reward g (Eq. 2 sums over N arms), while the right-hand side is derived as a per-arm quantity ("the contribution to the per-arm average gap is KRmax/N"). As written the inequality is dimensionally inconsistent; normalising both sides turns the rounding term into an O(K) total loss, which is fine but is not what the theorem states.
(d) Proposition 6. The argument "reducing wk to μfk removes this positive term without affecting the per-arm subproblems (which depend on wk/fk, not Bk)" is wrong: the per-arm subproblems depend on wk through the penalty (wk/fk)ai in Eq. (19), so lowering wk changes them. The conclusion wk⋆≤μ⋆fk can be rescued (the dual is non-decreasing in wk beyond μfk because ∣∂q/∂wk∣≤N while the penalty slope is N), but the "complementary slackness" step then only holds for the convex relaxation, not for the integer, hard-constraint bilevel problem the proposition is stated for. More importantly, the result is immediate: once both constraints are relaxed to expectations, E[∥a∥1∣k]≤Bk and ∑kfkBk≤Btot collapse to the single constraint ∑kfkE[∥a∥1∣k]≤Btot, so the per-context variables Bk are free and only one multiplier μ survives. The "two-phase" structure disappears entirely in the relaxation, which also means the LP bound used throughout the experiments is just the standard single-budget relaxation. The claim that this "has no analogue in single-budget RMABs" is therefore misleading.
This is consequential because, without Theorem 7, the paper has no guarantee connecting the tractable classes (FLag, MDVI) to the optimum, and the headline "vanishing fraction of the optimum" claim in the abstract is unsupported.
2. The remaining theoretical claims are folklore, and the experiments neither match the theory's assumptions nor test the paper's stated contribution; the reported numbers are internally inconsistent.
(a) Theorem 5 is the standard symmetry reduction. That an MDP invariant under a permutation group admits an optimal policy measurable with respect to the orbits (here, the counts n) is a textbook lumpability argument; Lemma 4 is exactly the usual proof. Calling the 112,000-instance check a "validation" of this theorem is tautological: the 100% rate is forced by symmetry (any violation would be a tie-breaking artefact), and the statement "we do not require indexability, UGAP, non-degeneracy" compares a symmetry reduction to asymptotic-optimality results that answer a different question. Note also that the restatement of Theorem 5 in A.3 ("the optimal two-phase policy is in Fagg") should read "there exists an optimal policy in Fagg"; with ties, not every optimal policy is in it.
(b) The portfolio effect is the known finite-N suboptimality of index policies. That fixed-priority (Whittle/LP-index) policies are not optimal for finite N and only asymptotically optimal is the starting point of Weber and Weiss (1990) and the entire subsequent literature; Proposition 2 re-establishes it numerically on a 3-arm instance. The Section 3.2 "counterexample" does not even exhibit a reversal: preferring a disengaged organic over an engaged churner is a single fixed ranking (org,0)≻(churn,1), so the described optimal policy is in Fidx. Remark 3's claim that the effect "requires N≥3" is supported only by a resolution-10 grid; at N=2, T=2 there do exist policies not representable by any fixed ranking. For instance, pulling type 0 when both arms share a state and type 1 otherwise induces, on the (type, state) categories, the cycle (0,0)≻(1,0)≻(0,1)≻(1,1)≻(0,0), so the claim is non-trivial and remains unproven.
(c) The theory does not apply to the experimental instances. Theorems 5 and 7 assume exchangeable types, but the benchmark arms are heterogeneous: each organic arm has its own distance-dependent pik and idle retention drawn per arm (C.1, C.3). The O(1/N) reference line in Figure 2 has no theoretical basis for these instances, and the MDVI sequence 6.2, 1.2, 3.2, 0.0, 1.1, 0.0 (Table 7) does not "track" it.
(d) The two-phase budget allocation is never exercised. All main experiments use fk=1/K, and the paper states (C.1) that the LP-optimal allocation then coincides with the uniform split, so "all performance differences arise from the arm selection policy". The only non-uniform experiment is two sentences in C.3 (3.9% to 2.8%) with no table. The paper's motivating contribution, deciding how to split budget across contexts, is thus untested.
(e) Text and tables contradict each other. "MDVI leads at most configurations" (6.3) and "achieves the lowest gap at every configuration" (Fig. 5 caption) versus Table 2, where NIP is better in six of eleven rows and Mitosis ties at N=500; abstract "0.0–3.2% gap" versus 6.4% at (20,2,4) and 6.2% at (50,3,5); "Mitosis ... 1.3% at N=500" (stated twice) versus 0.0% in Tables 2 and 7; "Fdual achieves 0.01–0.22% gap" versus 0.3–1.5% in Table 3, with standard errors of 1–3% that swamp every difference the "three clear patterns" are built on; L=500 iterations (6.3) versus 150 (C.2); "N=2,000 arms in 10 minutes" (abstract) versus 55 minutes (6.5); "four adversarial generators" (Table 6) versus five (Appendix D); "MDVI (Fagg)" in 6.3 versus FLag everywhere else; Table 3 labels Fdual as Fagg while the text calls it the smallest class; immediate rewards "R∈[0.8,1.2]" (6.1) versus pik∈[0.1,0.9] and churner rewards in [0.05,0.15] or [0.80,0.95] (C.3). Several of these are not typos but opposite claims about the same table.
This is consequential because, after removing the folklore and the unproven bound, what remains is a Lagrangian LP-index heuristic with subgradient price updates evaluated on instances where the novel part of the problem is switched off, reported with numbers that cannot all be correct.