{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"layer_decay_and_expiration_management__film_media_production","considered_ids":["H1","H2","H5"],"selected_id":"H1","dispositions":[{"hypothesis_id":"H1","lane":"FINALIST","reason":"Despite substantial collision with version-scoped comments and manual lifecycle states, the critic identifies a narrow, material, and directly falsifiable residual: revision-triggered expiry of prior-cut notes, explicit renewal, and preserved successor lineage. The stale-feedback problem is recognizable and moderately supported by direct product documentation."},{"hypothesis_id":"H2","lane":"RESEARCH","reason":"The orphan-cleanup problem is strongly supported, but the remaining claim bundles exhaustive cross-system dependency proof with impact-derived quarantine duration. Existing dependency discovery, orphan classification, and reversible purge components create substantial collision, while completeness across heterogeneous production systems remains insufficiently evidenced."},{"hypothesis_id":"H5","lane":"REJECT","reason":"Motion-picture archive restoration and periodically tested disaster recovery are already directly documented. The remaining cohort-sampling and whole-package drill specification is a procedural refinement amid very high overlap, and the critic recommends rejection."}],"selection_reason":"H1 offers the clearest differentiated opportunity: a bounded state-transition rule tied to promotion of a new cut, with explicit observable behavior and a straightforward falsifier. Its residual is narrower and more testable than H2's cross-system completeness claim, while H5 is too close to established archival recovery practice."}