Route recovery proof docs through backend wrapper

docs/release-control/v6/internal/subsystems/deployment-installability.md

@@ -214,6 +214,11 @@ verify, supervise, or inspect the routine local dev runtime, it must route them
 to the repo-root `npm run dev`, `npm run dev:verify`, and `npm run dev:logs`
 wrappers instead of teaching direct raw script invocations for those day-to-day
 flows.
+That same wrapper rule also applies to the managed recovery-proof docs in
+`tests/integration/README.md`: when those instructions tell operators how to
+bounce or verify the local managed runtime, they must use the repo-root wrapper
+surface such as `npm run dev:backend-restart` instead of documenting raw
+launcher commands directly.
 That same dev-runtime helper boundary also includes the auxiliary operator
 controls that start, stop, restart, or recover local development. The repo-root
 Makefile targets, `scripts/toggle-mock.sh`, and `scripts/clean-mock-alerts.sh` must

scripts/tests/test-hot-dev-bg.sh

@@ -12,6 +12,7 @@ PACKAGE_JSON="${ROOT_DIR}/package.json"
 FRONTEND_PACKAGE_JSON="${ROOT_DIR}/frontend-modern/package.json"
 DEV_LAUNCHD_WRAPPER="${ROOT_DIR}/scripts/dev-launchd-wrapper.sh"
 DEV_LAUNCHD_SETUP="${ROOT_DIR}/scripts/dev-launchd-setup.sh"
+INTEGRATION_README="${ROOT_DIR}/tests/integration/README.md"
 
 if [[ ! -x "${HOT_DEV_BG}" ]]; then
   echo "hot-dev-bg.sh not found or not executable at ${HOT_DEV_BG}" >&2
@@ -332,6 +333,14 @@ test_hot_dev_bg_script_advertises_managed_entrypoint() {
   assert_contains "hot-dev-bg routes launchd supervision guidance to managed wrapper" "${output}" "Rerun with: npm run dev"
 }
 
+test_integration_readme_uses_managed_backend_restart_wrapper() {
+  local output
+  output="$(sed -n '132,150p' "${INTEGRATION_README}")"
+
+  assert_contains "integration readme documents managed backend restart wrapper" "${output}" "npm run dev:backend-restart"
+  assert_not_contains "integration readme no longer documents raw backend restart script" "${output}" "./scripts/hot-dev-bg.sh backend-restart"
+}
+
 test_clean_mock_alerts_prefers_managed_runtime() {
   local test_dir fake_bin alert_history fake_hot_dev_bg action_log output
   test_dir="$(mktemp -d)"
@@ -563,6 +572,7 @@ main() {
   test_frontend_package_exposes_managed_runtime_entrypoints
   test_hot_dev_script_advertises_foreground_escape_hatch
   test_hot_dev_bg_script_advertises_managed_entrypoint
+  test_integration_readme_uses_managed_backend_restart_wrapper
   test_clean_mock_alerts_prefers_managed_runtime
   test_dev_check_uses_managed_runtime_status
   test_backend_restart_requires_managed_runtime

tests/integration/README.md

@@ -140,7 +140,7 @@ This mode attaches Playwright to the canonical dev browser entrypoint on `http:/
 
 If the managed runtime is not already running, the harness starts it. If you need the harness to reclaim existing unmanaged `5173`/`7655` listeners first, add `PULSE_E2E_HOT_DEV_TAKEOVER=1`.
 
-The managed recovery proof bounces the real backend through `./scripts/hot-dev-bg.sh backend-restart` and verifies that the browser shell reports the outage and then recovers through the proxy. When the harness attaches to an already-running managed dev session, `posttest` leaves that runtime running.
+The managed recovery proof bounces the real backend through `npm run dev:backend-restart` and verifies that the browser shell reports the outage and then recovers through the proxy. When the harness attaches to an already-running managed dev session, `posttest` leaves that runtime running.
 
 For deterministic paid-feature runs against an existing instance, provide one of:
 - `PULSE_E2E_BILLING_STATE_PATH=/absolute/path/to/billing.json` to let the harness write the billing state file directly.