Imagine you're trying to reconstruct a photograph from a mosaic of tiny color samples. You have enough sample points to uniquely identify the image — in principle, there's only one photo that could produce exactly those samples. But here's the catch: your samples are noisy, and some arrangements of sample points amplify that noise so catastrophically that your reconstruction is useless garbage. You need the samples to not just be sufficient in number, but geometrically well-arranged — stable under perturbation. That's the core mechanism of this paper, applied to quantum states instead of photographs. The committed claim: informational completeness — the standard criterion for whether a quantum measurement can uniquely identify a state — is necessary but not sufficient for stable reconstruction from finite data. The real dividing line is whether the POVM effects form a measurement frame in a σ-regularized operator geometry. Below that line, your estimators have divergent variance and your reconstruction is statistically meaningless no matter how much data you collect. This matters because continuous-variable (CV) quantum systems — think optical modes, mechanical oscillators, anything with infinite-dimensional Hilbert spaces — are the backbone of quantum communication and sensing. The field has been using informational completeness as the go/no-go criterion for measurement design for decades. This paper says that criterion is incomplete: you can have a measurement that uniquely identifies every state in theory but produces garbage estimates in practice. The framework introduces three regimes: inaccessible observables (cannot be estimated at all), weakly reconstructible ones (unbiased estimators exist but have infinite variance), and stably reconstructible ones (finite-variance unbiased estimators exist). The regime is determined by the range of what the authors call the POVM synthesis operator. The architecture is pure mathematical physics — no algorithms, no hardware, no simulations. This is operator theory and frame theory applied to quantum estimation. The reference state σ encodes prior information and defines the inner-product geometry in which frames are analyzed. The framework generalizes across homodyne tomography, heterodyne detection, photon-number-resolving measurements, and classical shadow protocols. It's a unification paper, not a benchmarking paper. The most striking result is the operational reinterpretation of quasiprobability distributions, particularly the Glauber-Sudarshan P representation. The P function's notorious singularities — the delta-function derivatives that make it not a proper probability distribution — are recast as a direct consequence of the measurement lacking a lower frame bound. The singularities aren't mathematical curiosities; they're telling you your measurement is unstable for the observable you're trying to estimate. This gives experimentalists a concrete diagnostic: if your quasiprobability representation blows up, your estimator variance is blowing up too. The validation is mathematical proof, which is both the strongest and most limited form of integrity. There's no cherry-picking risk — the theorems either hold or they don't. But there's also no experimental demonstration that the framework changes outcomes in practice. The paper is 60 pages of proofs with very few figures, explicitly soliciting community feedback. The predecessor landscape includes Holevo's quantum estimation theory, the classical shadow framework of Huang-Kueng-Preskill (2020), and the frame-theoretic approach to finite-dimensional tomography. This paper's contribution is extending frame theory to infinite-dimensional CV systems with the σ-regularization trick. The obvious next experiment the authors did not run: applying this framework to a specific CV measurement setup (say, heterodyne detection on a squeezed state) and showing numerically that the frame-bound criterion predicts estimator failure where informational completeness alone would not. My read: this is (c) — they're saving it. The 60-page theory paper establishes the framework; the companion paper with worked examples and numerics is the natural follow-up. Eight authors across four institutions suggests the collaboration has the bandwidth.
