Imagine you're in a dark warehouse full of identical-looking boxes, and you have a flashlight. If you shine it straight down from above, every box looks the same — you learn nothing about which boxes are stacked together. But if you tilt the flashlight at the right angle, shadows reveal the groupings instantly. This paper is about finding that angle for quantum measurements on non-abelian groups — the algebraic structures that encode graph isomorphism and lattice-based cryptography. The committed claim: for q-hedral groups Zq ⋊ Zp with p ≤ 7, four distinct Fourier sampling strategies were implemented on real trapped-ion hardware (IonQ Forte), and the theoretical hierarchy predicted by Moore, Rockmore, Russell and Schulman — that weak sampling fails, adapted-basis strong sampling works for large q, adjacent-pair sampling works for small q, and forgetful abelian sampling saturates below unit success — survives noise. At p=7, q=6, the quantum device exceeds the success probability of an exhaustively optimised classical algorithm for up to three oracle queries. The ladder matters here. The classical baseline is not a throwaway straw man — it is an exhaustively optimised brute-force strategy over all possible classical measurement outcomes. For up to three oracle queries at p=7, q=6, the quantum approach wins. That said, these are small instances (primes up to 7), and the gap is modest. Nobody is breaking lattice cryptography today. The result is a proof-of-concept that the theoretical measurement-basis hierarchy is physically real, not a simulation artifact. Architecturally, this belongs to the non-variational query-complexity family of quantum algorithms. No VQE, no QAOA, no parameterized circuits to optimize. The circuits implement exact group-theoretic Fourier transforms — the quantum Fourier transform generalized to non-abelian groups — followed by measurements in carefully chosen bases. The hardware dependency is on the high connectivity and relatively low gate error of IonQ Forte's trapped-ion architecture, which allows the multi-qubit entangling operations these Fourier transforms require without deep swap networks. Integrity is solid for what this is. The validation is experimental, not simulated: actual counts from a real quantum device, compared to theoretical predictions and to a classical algorithm whose performance was computed exhaustively. The four sampling strategies tested are not cherry-picked — they come directly from the Moore-Rockmore-Russell-Schulman theoretical framework established over two decades ago. The paper is honest that weak sampling fails (as predicted), forgetful sampling saturates (as predicted), and only the right strong sampling basis succeeds. There is no cherry-picking of the easy cases. The milestone question is where this gets interesting — and sobering. The q-hedral groups tested here are toy-sized (p ≤ 7 means 3-qubit registers). The graph isomorphism and cryptography applications require scaling to groups where p is exponentially large. The next concrete milestone would be demonstrating the adapted-basis advantage at p ≥ 13 (requiring 4+ qubit coset registers), where the circuit depth grows and noise becomes more punishing. That is likely 2-3 years out given current hardware trajectories. The experiment not run: scaling to p = 11 or 13. At p = 11 you need 4-qubit registers and deeper circuits, and gate errors would likely erode the quantum advantage. The honest read is (a) — they ran out of hardware fidelity budget. The IonQ Forte can handle the 3-qubit registers for p ≤ 7, but the next prime up would require significantly deeper circuits where decoherence dominates. This is not a criticism; it is the reality of 2025-era quantum hardware. The paper wisely stops where the signal is still clean.
