feat(rate-of-closure): add responsive wind strategy workflow

.github/workflows/maturin-swing-core.yml

@@ -110,6 +110,9 @@ jobs:
       - name: Run Rust<->Python parity suite (gate, non-skipped)
         env:
           PYTHONPATH: src
+          # This Rust-only lane must not import unrelated plugins cached on a
+          # self-hosted runner (for example pytest-qt without PyQt6 installed).
+          PYTEST_DISABLE_PLUGIN_AUTOLOAD: "1"
         run: |
           python -m pytest \
             src/shared/python/swing_sim/tests/test_parity_rust.py \

docs/release/RATE_OF_CLOSURE_CAMPAIGN_MANIFEST.md

@@ -0,0 +1,70 @@
+# Rate of Closure Campaign Manifest
+
+`rate_of_closure_campaign.v1.json` is the machine-readable authority for
+implementation and release claims in the Rate of Closure, impact, and flight
+campaign. It complements the issue tracker and handoff; it does not replace
+either one.
+
+## What It Records
+
+The manifest normalizes metadata that was previously repeated in narrative
+handoffs:
+
+- one record for every primary campaign epic, program, or release gate;
+- canonical repository specifications and GitHub issue authorities;
+- shared pull-request carriers and exact observed branch heads;
+- all four supported product surfaces;
+- commit-bound local or hosted test evidence;
+- explicit evidence gaps, limitations, dependencies, and child issue IDs;
+- release status and the exact `main` SHA when a release is eventually proven.
+
+Program delivery uses four mutually exclusive stages:
+
+1. `specified_only` — acceptance scope exists, but no implementation carrier is
+   claimed;
+2. `implemented_on_feature_stack` — implementation exists outside a qualified
+   protected release;
+3. `protected_merged_to_parent` — a parent merge and passing protected checks
+   are both recorded;
+4. `released_to_main` — the program release SHA matches a merged carrier into
+   `main`.
+
+A feature-branch merge is not automatically stage 3. A local test pass is not
+a protected merge, and a protected parent merge is not automatically a
+default-branch release.
+
+## Validation
+
+Run the deterministic validator from the repository root:
+
+```powershell
+python scripts/rate_campaign_manifest.py
+```
+
+Emit the versioned JSON Schema generated from the same Pydantic models:
+
+```powershell
+python scripts/rate_campaign_manifest.py --schema
+```
+
+The validator fails closed when it finds undeclared fields, missing programs,
+duplicate IDs, malformed SHAs, unresolved carrier or evidence references,
+missing repository paths, placeholders, or contradictory delivery/release
+states.
+
+## Maintenance Procedure
+
+Update the manifest in the same commit as a material carrier, test-evidence, or
+release-state change:
+
+1. Refresh only the affected carrier head, state, and merge evidence.
+2. Add commit-bound test evidence; do not replace failed or queued evidence
+   with narrative optimism.
+3. Advance a program stage only when its validator preconditions are true.
+4. Record limitations and missing evidence even when implementation exists.
+5. Run the focused manifest tests and validator.
+6. Update the root and Rate campaign handoffs in the same commit.
+
+The checked-in snapshot does not query GitHub at validation time. This keeps CI
+deterministic and avoids API-rate coupling, but means a maintainer must refresh
+carrier observations deliberately after a merge, push, or protected run.