Truth Before Ontology: Apermanent Structural Targets and the Meta-Level Limits of the Scientific Realism Debate

The quantum wavefunction has been debated for a century without convergence not because philosophers are insufficiently clever, but because the debate has been attempting classification of a target that is not yet in a condition to be classified. Dark matter is the same situation at cosmological scale: four independent lines of evidence converge on a structural role that nothing with a substrate-stable identity has been found to fill. These are not hard cases for scientific realism. They are the debate's most precise results—accurate expressions of what tracking looks like when a target resists ontological completion while continuing to constrain everything inquiry does with it. This paper argues that the scientific realism debate's persistent instability has a structural source that no argument within the debate can address: it begins too late, presupposing science as a first given and truth as primarily ontological, while the conditions under which scientific targets remain answerable have already been decided before the ontological question is raised. Many of the debate's most recalcitrant targets are neither stable substances nor merely instrumental constructions. Their scientific reality lies not in substrate-stable identity but in repeatable re-enterability, constraint-bearing role, and resistance to arbitrary revision across transformations of model, context, and realization. I call these apermanent structural targets. They do not merely resist classification; they generate the debate's constitutive antinomy by simultaneously activating the realist demand for answerability and the anti-realist demand for revisability. The debate has not failed to resolve this antinomy because the arguments are bad. It has failed because the targets themselves make resolution at the first-order level structurally unavailable. On this basis, I distinguish ontological truth within science from the existential truth of science itself—whether science still stands in a mode of occurrence open to being revised by reality rather than merely processing what its current admissible interfaces allow through. Nine familiar repair strategies, from structural realism to stance pluralization, can be read as attempts to protect one of three legitimacy conditions—object-answerability, process-fallback, debate-sustainability—by displacing load onto another. None preserves all three simultaneously. The paper therefore proposes neither a new mandatory doctrine nor an anti-realist retreat, but a permissive tracking-stance: a methodological permission to hold the diagnostic question open at a pre-classificatory level, without forcing an immediate choice between heavy ontology and instrumental deflation. At most, the paper offers a partial fallback. The wavefunction's century of unresolved interpretation and dark matter's four decades without a substrate identity are not embarrassments to be eventually overcome. They are what the debate looks like when it is working correctly at the right level.