Fail-closed reproducibility for finite quantum sequential growth: exact controls, independent replay, and resource boundaries

We report a fail-closed reproducibility framework for finite quantum sequential-growth calculations. The framework separates candidate evidence from controls, binds claims to versioned machine-readable artifacts, requires independent replay before promotion, and records resource limits as open boundaries. Existing bounded artifacts include a five-seed small-n sampler regression, 40 deterministic sampler fixtures, eight four-dimensional Minkowski-control fixtures, and five synthetic supervisor modes. In a separate exact random-growth scaling review, primary and independent paths complete growth through cardinality 38 with identical digests; an attempted step from source size 38 detects at least 4,097 down-sets and fails closed at the declared cap of 4,096. This is a resource-limit observation, not an impossibility theorem. No candidate production evidence, continuum-limit result, or final-theory claim is made.