Imagine you run a call center and you want to know: how much extra information can my best operator extract from a caller compared to a standardized script? The "extra" depends on the diversity of caller types — if every caller sounds identical, no operator skill helps. This paper does the same thing for quantum communication: it measures the diversity (coherence) of a sender's ensemble of quantum states and proves that diversity caps the advantage any clever measurement can extract. The committed claim: the continuous ensemble of pure states distributed under Haar measure is maximally coherent in every finite dimension, achieving the tight upper bound on basis-free relative entropy ensemble coherence. This is a clean mathematical result — a supremum identified and proven tight — not a computational or experimental advance. It settles a question that was open in the resource-theoretic coherence literature: what is the ceiling? The operational payoff comes in two theorems. First, ensemble coherence bounds the gap between what a postselective (probabilistic) measurement can extract versus the Holevo quantity for two equiprobable pure states. In plain terms: coherence quantifies how much "extra juice" you can squeeze by accepting measurement outcomes probabilistically rather than deterministically. Second, for the qubit B92 quantum key distribution protocol, the coherence of Alice's ensemble upper-bounds the asymptotic secret key rate. This gives coherence a concrete cryptographic meaning — it caps how many secret bits per channel use the protocol can deliver. The architecture is pure resource theory — no algorithms, no hardware, no simulations. The tools are information-theoretic: relative entropy, Haar integration, Holevo bounds. The paper lives in the mathematical physics tradition of quantifying quantum resources (entanglement, coherence, magic) and proving tight bounds. It leans on the basis-free coherence framework introduced by Åberg and extended by Streltsov, Chitambar, and others. Integrity is straightforward for a theory paper: the results are mathematical proofs, not empirical claims. There is no cherry-picking risk in the usual sense — either the bound is tight or it is not. The paper proves tightness by construction (the Haar ensemble achieves the bound). The scope limitation is that the operational results are proven only for specific settings: two equiprobable pure states (measurement advantage) and qubit B92 (key rate). Generalizations to arbitrary ensembles or protocols remain open. The milestone question for this line of work is whether ensemble coherence can bound key rates for more practical QKD protocols — BB84 with decoy states, measurement-device-independent QKD, or continuous-variable protocols. The B92 result is a proof of concept: the simplest protocol in the simplest dimension. The field needs the same inequality for protocols people actually deploy. The obvious next experiment not run: extending the key-rate bound beyond qubit B92 to BB84 or higher-dimensional protocols. The honest read is (a) — the mathematical machinery for multi-dimensional or multi-basis protocols is substantially harder, and this paper establishes the principle in the simplest case first. There is no reason to suspect the result fails; there is reason to suspect the proof technique does not generalize easily.
