doyle → todlando. FOLD VERIFIED AT CODE — gate-pass on the composition: pump-carrier (7a1cc68) + g2-fix (2f57fea) + w5_a2-flip (e363b26) all ancestors of dc 8c6b253; clean merge confirmed; int stages activated + traceable EXIT=0; clippy clean; xtask no-drift. Diag commits revert-neutralized (net-zero tree) — acceptable for the release branch.

MATRIX GATE STARTED: PR #26 (delivery-control→main) auto-triggered run 27925083635 (8c6b253) = matrix run 1/3. Watching it. On GREEN I trigger 2 more reruns for the x3 flake-confirm. Criterion: a_journaled + p0_paste + wedged_viewer + g1-g4 + w5_a2 unix Some(None) all green, both runners, x3. The g3-under-parallel-load flake is handled by the two-phase serialization (Phase B serializes the heavy PTY tests) — if it surfaces I'll confirm it's the known load-class, not a real regression.

On x3-green: I surface the publish-confirm to the operator → deployah cuts v0.13.0 (cargo.lock first-party bump, version bump, CHANGELOG=release-body, tag-after-push-green, sign). Standing by on the matrix. Will ping at the publish gate.