Hilbert Event Geometry and the Quadratic Weight from Operational Primitives: A Conditional Reconstruction with an Assumption-Dependence and Failure Map

We study how much of the Hilbert-space and Born-rule structure of quantum theory can be obtained from operational primitives that involve records but not probabilities. The setting is a regime in which coherent alternatives recombine without leaving a stable record. We obtain three results of different kinds. Geometry. From composable contributions, record-inert reversible controls and amplitude scalings, under named conditions, we obtain a real inner-product event geometry. The existence of the inner product is equivalent to bounded mixing and is assumed in that form. In a worked interferometer with one mode per channel, the metric is not an input: metric-free port swaps and phase flips fix the form up to scale, and boundedness only forces the other controls to preserve it. Its uniqueness up to scale, the orthogonality of distinct outcomes, and the representation of events by orthogonal projections are derived. Obstruction. This geometry, even with full symmetry about the preparation and exact dyadic standards, does not select the quadratic weight: the preparation mixed with the maximally mixed state survives. The structural operations exhibited here certify only values on which the quadratic weight equals the counting weight. On restricted event families where Gleason’s theorem is unavailable, a calibration order oriented by the quadratic magnitude selects it: given exact standards, a state is the quadratic one exactly when it respects all such brackets. Whether such an orientation can be earned is the open problem. Given such brackets and refinement only to a finite depth K, every admissible state lies within 2^-K of the quadratic weight, without the Archimedean step. On the full projection lattice, Gleason’s theorem with null support, or sphere-transitive symmetry with null support, gives other conditional routes. Bridge. A predictive probability kernel is introduced separately. Its identification with the real quadratic weight ||P_A ψ||² / ||ψ||² is conditional on kernel-admission hypotheses, one of which is equivalent to the Born value; the complex form needs phase invariance in addition. The hypotheses select a class of contexts, and none is shown to be forced. For each we state the physical feature it expresses and a sufficient condition where one is known, and we work out one apparatus in which the geometry-stage hypotheses hold together. For all but one we give a context in which it fails (for the founding conditions, one that leaves the regime); for the remaining one, TCC-1, we give an illustration. This paper is the first in a series built on a common framework, Structural Ledger Consistency (in preparation).