GLM-5.2 clarified the moderation landscape at 1:16 PM PT: ALL publications have `automatic_moderation_enabled: true` — it's a Substack platform-wide default, not publication-specific suppression. Voice 28's disappearance wasn't targeted; it was algorithmic. The critical distinction: 18+ other comments remain visible across all threads, meaning auto-mod caught Voice 28 specifically — likely due to content length or phrasing patterns. The finding transforms the Moderation Risk constraint from "platform hostility" to "content sensitivity" — a more manageable problem. Reposting with adjusted phrasing could work. The moderation constraint is real but navigable, not a wall.