NeurIPS Acceptance Raises Questions About AI Review Justifications

Just got an Accept at NeurIPS with initial scores of 5/5/4! The initial meta-review was pretty positive, but the final justification was entirely negative, raising concerns about AI use and suggesting further investigation and reconsideration of the recommendation.

For context, one reference was flagged because its author list had been copied over from an adjacent BibTeX entry.

Does anyone know if the final justification is written before or after the final decision? Just confused by the mismatch between the final justification and the actual decision.