The Sun Won't Confess: Three Probes, Two Compositions, No Closure

The solar abundance problem is a defect with three independent witnesses. Standard solar models built on the modern spectroscopic composition (low metallicity, AGSS09) fail the helioseismic observables: convection-zone depth R_CZ = 0.7223 ± 0.0053 against the seismic 0.713 ± 0.001 (defect +1.7σ with model errors charged whole; +9σ if they are not), surface helium Y_S = 0.2317 ± 0.0059 against 0.2485 ± 0.0035 (−2.4σ), and a sound-speed profile with a localized discrepancy of order 1% below the convection zone. Models built on the older composition (high metallicity, GS98) pass every seismic observable (R_CZ defect −0.29σ, Y_S −0.86σ, sound speed flat at a few parts in 10³) — but the photospheric spectroscopy that produced the low composition is direct measurement, not a model preference, and it has only strengthened with 3D non-LTE technique. The third witness, neutrinos, leans toward the high composition: Borexino's final CNO flux, Φ_CNO = 6.7 (+1.2/−0.8)×10⁸ cm⁻² s⁻¹, combined with its ⁷Be and ⁸B fluxes, disfavors the low-metallicity B16 model at 3.2σ — yet the measured CNO flux sits above both compositions' predictions, and the prediction itself moved ~15% in 2025 when the CNO cycle's slowest reaction, ¹⁴N(p,γ)¹⁵O, was remeasured upward. A different recent model generation flips the ⁸B preference to the low composition entirely. Verdict: the solar core is not closed. Helioseismology convicts the spectroscopic composition; spectroscopy refuses to recant; neutrinos lean with seismology but cannot yet testify at the precision the case requires, because their key prediction is a nuclear cross-section that is still being re-measured. The cheapest exits are priced in §4 — an opacity revision at the convection-zone base, the nuclear-rate settlement, and surface–core decoupling by early accretion — and none is endorsed. The best-measured star in the sky keeps its secret, and this paper's business is the size of the lock, not a story about the key.