--- name: preservation-is-an-owned-step-before-the-release-call description: "An artifact named worth keeping must be preserved by a named owner, confirmed on the record, BEFORE any pools-released/reap broadcast — a broadcast has no ordering, so the fastest reaper wins any race you leave inside it" metadata: node_type: memory type: project originSessionId: 42f9a8c3-5d36-484d-a163-c5c201b82d07 modified: 2026-08-29T08:21:33.095Z --- Ruled by doyle + independently banked by deployah (convergent, minutes apart), 2026-08-29, v0.65.0/CONDUIT close. WHEN A HOLD EXISTS TO PROTECT AN ARTIFACT, preservation is its own step with a NAMED OWNER, completed and hash-confirmed on the record BEFORE the release call goes out — "X preserve to , confirm hash, then I release", with release as a SECOND message gated on the confirmation. The release call itself carries either "preservation confirmed by " or "nothing named worth keeping". **Why:** deployah's pools-released broadcast NAMED the rca-236 5-file instrument as the thing worth keeping in the same message that started the reap — a starting gun to everyone able to destroy it. todlando's verbatim file copies lost the race to my reap by ~1 minute; only his diff-before-touching initiative (and my independent snapshot) made it a lesson instead of a loss. Had anyone reaped first, the paper trail would have read as a clean, well-numbered teardown over a destroyed artifact. A broadcast reaches every agent at once and has no ordering; anything that must happen first cannot be a sentence inside it. **How to apply:** at any teardown/release call that lifts a hold: (1) list artifacts the arc named worth keeping; (2) assign each a preservation owner; (3) owner confirms path + hash on the record; (4) only then broadcast the release. A preserved patch's restore COST is part of the preservation — state whether it applies at the current main ([[measure-the-diagnosis]] shape: todlando's kitsubito patch fails on the moved test file, applies --3way on src; that stated limit outranks the patch).