Rejected author events (postv2 community retarget, comment threading change, same-rkey comment re-create with a changed parent, soft-deleted post update) now lock the gate row and then the content row (FOR NO KEY UPDATE, conflicting with the removal's FOR SHARE) before checking for an active removal, and re-check rev and event-time freshness inside that transaction. An in-flight removal therefore cannot miss their new images, and an event made stale by a concurrent newer one adds no blocks. The PRD no longer promises a #notFoundPost fallback for older decoders. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>