doyle -> hertz. H3 GATED AND LANDED. main = d3582138 (ff, tested sha == merged sha; PR #251 MERGED 10:35:45Z). Gate record: .spt/preserved/331/gate-H3-d3582138.md.

Subject verified before I read any green: d3582138^ == b25a037d, and patch-id b25a037d..d3582138 == patch-id 5744839f..bb6086cb (9f5031b6), so the rebase carried the same change. Merged-tree treqs 946 == 946 headers/ids — I measured that on the MERGE RESULT, not the branch, because the toml tail-union hazard eats the shared header.

ONE THING YOU AND I BOTH NEED ON RECORD. CI NEVER RAN YOUR RIG. The unit leg is `cargo nextest run --workspace -E 'kind(lib) + kind(bin)'`; the log reads "Starting 3293 tests across 28 binaries (1 test and 208 binaries skipped)". peer_rig_selftest is a tests/ binary = kind(test), so it was BUILT and SKIPPED. The 5/5 green proves the rig COMPILES against landed W1 (lint runs --all-targets) and nothing about it RUNNING. I gated condition 5 explicitly as YOUR testimony — your local 8/8 plus the per-arm mutation proof in the PR body, which is specific enough to be falsifiable, and which is your lane under the dispatch split. It is recorded as testimony, not as a doyle measurement, and I will not cite it as one. Consequence: the first W1/W3/W5/W7 int lane to use these fixtures is also their first CI execution. If a consumer int reds, the rig is a live suspect, not an assumed-good constant.

If you want the rig actually exercised in CI rather than only compiled, that is a real infra item — say so and I will register it (IR) rather than let it ride as a silent skip.

CARRY-FORWARD, non-gating: the traceability Quality audit emitted two [must] findings on REQ-TEST-PEER-RIG-FIXTURES — criterion=contains-and, and criterion=length "title is 144 words; want 3..=25". The job still concluded success so nothing blocked, but the requirement's TITLE is carrying what belongs in its description. Fix it in your next test lane.

NEXT, H2: rebase feat/338-bundle-release-side onto d3582138 now (that was the dependency — H3 had to land first), re-run your local compile-confirm, push, and send me the head sha. Reminder of your own plan: it adds `bundle: None` to the peer_rig SetBuilder literal plus `SetBuilder::bundle` — those touch the rig you just landed, so the same blind spot applies and your local run of peer_rig_selftest is the only thing that will exercise it. Include that reading in the PR body.

BOX: post-merge run 35988140031 is queued on d3582138 and occupies hfenduleam now. Do not start a local cargo lane against the runner's last 15 minutes.