Nested Bubble/Gear (NBG-01) v0.3.1: E2-A Canonical Execution and Deterministic Replay

Nested Bubble/Gear (NBG-01) v0.3.1 reports the canonical execution and deterministic replay of Experiment E2-A: Interface-Aware Challenge, following the publicly timestamped NBG-01 v0.3.0 preregistration and pre-run freeze. E2-A is the first challenge experiment in the NBG-01 lineage in which predictive closure of the inexpensive observer hierarchy is not guaranteed by construction. The transition law introduces a preregistered interface dependence through designated microstate bits x3 and x4: q_A = (1 - kappa - delta)m_A + kappa m_B + delta x3 q_B = (1 - kappa - delta)m_B + kappa m_A + delta x4 with p_A = eta + (1 - 2 eta)q_A p_B = eta + (1 - 2 eta)q_B. The complete frozen grid contains 64 parameter points spanning four values each of eta, kappa, and delta. The 16 delta = 0 points form a baseline-continuity slice, while the remaining 48 positive-delta points constitute the interface-aware challenge. The canonical execution was performed without modifying the observer ladder, parameter grid, thresholds, witness rules, serialization convention, mixing definition, cost rules, or anti-rescue policy frozen in v0.3.0. All 16 baseline transition-kernel digests matched the NBG-01 v0.1.2 parent kernels, satisfying the preregistered continuity gate. The identity control O3 passed all 64 of 64 parameter points. Across the complete 64-point grid, the terminal promotion outcomes were: 9 PROMOTE_O0 14 PROMOTE_O_IF 41 FAIL-TRUST-SEPARATION The principal challenge result appears on the 48 positive-delta points. The inexpensive majority observer O0 and the weight-based observer O1 both fail predictive closure at 48 of 48 challenge points. In contrast, the interface-aware observer O_IF = (m_A, m_B, x3, x4) and the richer observer O2 = (w_A, w_B, x3, x4) retain exact one-step and two-step predictive closure and composition at 48 of 48 challenge points. O1 and O_IF each realize 16 effective classes, making their comparison especially informative. Although they have equal effective description size, only O_IF retains the interface variables that directly enter the modified transition law. E2-A therefore distinguishes observers by the relevance of the information retained rather than by observer size alone. Predictive closure does not automatically imply sufficient metastable persistence. Among the 48 positive-delta challenge points, O_IF passes the preregistered trust-separation gate at 14 points and fails it at 34 points. No challenge point requires promotion to the identity observer. All 256 majority-sector mixing analyses resolved within one to three discrete steps. The delta = 0 sectors retain rank-one structure, while positive-delta sectors have rank four under the frozen diagnostics. No near-rank-one ambiguity condition was triggered. Following the canonical execution, a clean same-host isolated deterministic replay was performed from the identical frozen source. Every declared deterministic scientific artifact reproduced byte-for-byte, yielding: REPLAY_EXACT The replay is a same-host deterministic reproduction and is not presented as an external independent-laboratory replication. The canonical execution result is: E2A_CANONICAL_EXECUTION_COMPLETE NBG-01 v0.3.1 does not establish a new physical law or make claims about spacetime, quantum mechanics, consciousness, or nature's literal ontology. It reports the outcome of a finite preregistered computational challenge testing predictive closure, observer adequacy, interface relevance, composition, metastable trust separation, and observer promotion under rules published before the result was known. This release closes the E2-A experiment. Any subsequent hidden-microstructure challenge will be preregistered separately rather than modifying or rescuing the frozen E2-A result.