Nested Bubble/Gear (NBG-01) v0.1.2: Canonical Execution and Deterministic Replay

Nested Bubble/Gear (NBG-01) v0.1.2 is the canonical execution and deterministic replay release following the protocol-repair and freeze publication NBG-01 v0.1.1. Experiment 1 was executed under the conventions frozen before canonical output was inspected. The experiment is designated POSITIVE_CONTROL; the numerical-zero convention is fixed at 1e-12; analytically rank-one killed sectors use an operational conditional mixing time of one discrete transition; continuous-hazard lifetime is the primary trust convention; geometric expected lifetime is retained as a robustness comparison; and the trust threshold is fixed at Θ ≥ 10 with no rescue band. The canonical execution reproduces the repaired Experiment 1 boundary established during protocol analysis. Observers O0, O1, and O2 exhibit exactly zero one-step and two-step predictive leakage across the complete 16-point parameter grid. The maximum raw composition defect is approximately 1.17 × 10^-15, below the frozen numerical-zero convention. All killed majority-sector kernels are rank one with zero within-sector row spread. Under the frozen trust gate, 9 of 16 parameter points receive PASS-CHEAP and 7 receive FAIL-TRUST-SEPARATION. The seven failed points are preserved rather than rescued or removed, demonstrating that exact predictive closure does not imply sufficient metastable persistence. Following the canonical execution, a clean same-host isolated replay was performed from the identical frozen source. Every deterministic scientific artifact reproduced byte-for-byte. The replay environment matched the canonical execution on all frozen comparison fields except intentionally varying execution metadata such as timestamp and execution label. This replay is a deterministic same-host reproduction and is not presented as an external independent-laboratory replication. NBG-01 v0.1.2 does not establish a new physical law or make claims about spacetime, quantum mechanics, consciousness, or nature's literal ontology. Its purpose is to close the v0.1.x Experiment 1 repair cycle by publishing the canonical execution, frozen receipts, deterministic replay, and custody evidence required for subsequent negative-control and challenge experiments.