Harden WP-0024 recovery execution gates

Assistant: codex
Assistant-Model: gpt-5.6-sol
Assistant-Session: 01a02669-87ee-7a31-b111-edc95a16e0fa

Makefile

@@ -476,10 +476,23 @@ audit-whitehat-e2-cleanup: ## Remove the exact engagement credentials and reload
 	python3 scripts/audit-core-whitehat-e2-credentials.py cleanup --confirm WH-ENG-20260822-AUDIT-E2-01:attended
 
 audit-database-lease-recovery-preflight: ## Read-only gate for the restart-free database lease exercise
-	python3 scripts/audit-core-recovery-preflight.py database-lease
+	python3 scripts/audit-core-recovery-preflight.py database-lease $(AUDIT_DATABASE_LEASE_PREFLIGHT_ARGS)
+
+audit-database-lease-approval-check: ## Validate an approved T02 receipt without mutation
+	@test -n "$(AUDIT_DATABASE_LEASE_APPROVAL)" || { echo 'set AUDIT_DATABASE_LEASE_APPROVAL=/path/to/approved-receipt.json' >&2; exit 2; }
+	python3 scripts/audit-core-database-lease-recovery.py validate-approval \
+	  --approval "$(AUDIT_DATABASE_LEASE_APPROVAL)"
+
+audit-database-lease-recovery-exercise: ## Run the separately approved attended T02 exercise
+	@test -n "$(AUDIT_DATABASE_LEASE_APPROVAL)" || { echo 'set AUDIT_DATABASE_LEASE_APPROVAL=/path/to/approved-receipt.json' >&2; exit 2; }
+	@test -n "$(AUDIT_DATABASE_LEASE_DRIVER)" || { echo 'set AUDIT_DATABASE_LEASE_DRIVER=/path/to/approved-load-driver' >&2; exit 2; }
+	python3 scripts/audit-core-database-lease-recovery.py exercise \
+	  --approval "$(AUDIT_DATABASE_LEASE_APPROVAL)" \
+	  --load-driver "$(AUDIT_DATABASE_LEASE_DRIVER)" \
+	  --confirm RAILIANCE-WP-0024-T02:attended
 
 railiance01-reboot-preflight: ## Read-only gate for the coordinated single-node reboot
-	python3 scripts/audit-core-recovery-preflight.py node-reboot
+	python3 scripts/audit-core-recovery-preflight.py node-reboot $(RAILIANCE01_REBOOT_PREFLIGHT_ARGS)
 
-.PHONY: audit-whitehat-e2-preflight audit-whitehat-e2-project audit-whitehat-e2-cleanup audit-database-lease-recovery-preflight railiance01-reboot-preflight apps-pg-isolation-probe
+.PHONY: audit-whitehat-e2-preflight audit-whitehat-e2-project audit-whitehat-e2-cleanup audit-database-lease-recovery-preflight audit-database-lease-approval-check audit-database-lease-recovery-exercise railiance01-reboot-preflight apps-pg-isolation-probe
 .PHONY: db-deploy db-status db-shell db-logs forgejo-db-deploy forgejo-db-status forgejo-db-shell apps-pg-deploy apps-pg-status apps-pg-shell apps-pg-logs net-kingdom-pg-inter-hub-networkpolicy-deploy pg-deploy pg-status pg-pgpool-check valkey-deploy valkey-status openbao-repo openbao-dry-run openbao-overlay-apply openbao-verify-login-overlay openbao-deploy openbao-status openbao-verify openbao-verify-post-unseal openbao-configure-initial openbao-configure-ssh openbao-verify-ssh openbao-verify-authenticated openbao-configure-external-secrets-issue-core openbao-configure-external-secrets-activity-core openbao-configure-external-secrets-forgejo openbao-validate-restore-evidence openbao-validate-emergency-evidence credential-grants-validate credential-change-applier-dry-run credential-change-applier-apply-plan credential-change-applier-apply credential-change-runbook credential-change-record-evidence credential-change-lifecycle-plan credential-change-lifecycle-event credential-change-import-inventory agent-high-risk-boundary-check openbao-credential-change-appliers-dry-run openbao-configure-credential-change-appliers openbao-token-grants-dry-run openbao-configure-token-grants openbao-verify-token-grants-dry-run openbao-verify-token-grants openbao-verify-token-grants-smoke credential-helper-dry-run credential-tests credential-exec-ops-warden-smoke argocd-bootstrap-dry-run argocd-bootstrap-deploy argocd-repo-apply argocd-status backup forgejo-backup forgejo-backup-dry-run forgejo-backup-status forgejo-package-prune forgejo-package-prune-dry-run reef-deployables apps-pg-evidence consumption-preflight help

docs/audit-core-database-lease-approval.example.json

@@ -0,0 +1,29 @@
+{
+  "procedure": "audit-core-database-lease-recovery",
+  "task_id": "RAILIANCE-WP-0024-T02",
+  "status": "pending",
+  "approval_id": null,
+  "window": {
+    "start": null,
+    "end": null
+  },
+  "abort_operator": null,
+  "owners": {
+    "audit-core": {
+      "acknowledged": false,
+      "message_id": null
+    },
+    "rapp-postgres": {
+      "acknowledged": false,
+      "message_id": null
+    },
+    "railiance-platform": {
+      "acknowledged": false,
+      "message_id": null
+    }
+  },
+  "synthetic_load": {
+    "contract_id": null,
+    "driver_revision": null
+  }
+}

docs/audit-core-database-lease-recovery.md

@@ -29,6 +29,25 @@ The result must have `automated_checks_passed: true`. It will remain
 provided. The helper reads Secret metadata and key names only; it never reads
 Secret data.
 
+The live harness is
+`scripts/audit-core-database-lease-recovery.py`. Start from the deliberately
+pending `docs/audit-core-database-lease-approval.example.json`; the actual
+value-safe receipt must name a ≤15-minute window, approval id, abort operator,
+State Hub acknowledgement ids from audit-core/rapp-postgres/platform, and an
+approved synthetic-load contract plus driver revision. Validate without
+mutation:
+
+```bash
+make audit-database-lease-approval-check \
+  AUDIT_DATABASE_LEASE_APPROVAL=/path/to/approved-receipt.json
+```
+
+The separately approved load driver is an executable that accepts `baseline`,
+`expect-unavailable`, `expect-recovered`, and `cleanup`, plus
+`--contract-id ID`. Each call returns only its exact value-safe JSON evidence
+shape. Unexpected keys—including a bearer accidentally returned as `token`—
+make the harness fail without echoing driver output.
+
 ## Attended sequence
 
 Keep one terminal attached from baseline through recovery. A value-handling
@@ -41,11 +60,14 @@ bodies and clear all transient JSON from memory or mode-`0600` temporary files.
 2. Start bounded retrying synthetic traffic. Its evidence records only request
    identifier, attempt time, response class, and final accepted/duplicate
    result—never the bearer or payload.
-3. Inside the attended process, read the current mounted username without
+3. Inside the attended process, list only lease handles below
-   emitting it. List only lease handles below
+   `sys/leases/lookup/database/creds/audit-core-runtime`. OpenBao's lease lookup
-   `sys/leases/lookup/database/creds/audit-core-runtime`, look each up, and
+   returns issue/expiry/TTL metadata but not the leased username, so exact
-   retain the one handle whose lease username matches. Abort unless exactly one
+   selection is deliberately stricter: there must be exactly one live handle
-   live handle matches. Clear every lookup response and username immediately.
+   under that exact runtime prefix, and its issue/expiry interval must contain
+   the current database Secret refresh time. Abort for database-owner
+   disambiguation if zero or multiple handles exist. Retain only a one-way
+   handle fingerprint in evidence and clear the lookup response immediately.
 4. Recheck that the Secret `resourceVersion`, pod UID, and selected handle have
    not changed. Revoke only that handle. Do not revoke the role prefix and do
    not revoke the ESO parent token.
@@ -64,6 +86,20 @@ bodies and clear all transient JSON from memory or mode-`0600` temporary files.
    the predecessor lease is revoked. The replacement remains managed by ESO
    and expires through its normal database role TTL.
 
+Only inside the approved window, with the receipt and reviewed driver:
+
+```bash
+make audit-database-lease-recovery-exercise \
+  AUDIT_DATABASE_LEASE_APPROVAL=/path/to/approved-receipt.json \
+  AUDIT_DATABASE_LEASE_DRIVER=/path/to/approved-load-driver
+```
+
+The exact confirmation string is embedded in the target. The harness rechecks
+the Secret/pod/lease immediately before revocation, requires at least five
+minutes of remaining lease TTL to avoid the normal ESO refresh race, performs
+best-effort forced reconciliation on an interrupted post-revocation run, and
+never issues a rollout/restart command.
+
 ## Time bounds and abort path
 
 - Planned outage observation: at most 120 seconds.

docs/openbao-reboot-snapshot-receipt.example.json

@@ -0,0 +1,22 @@
+{
+  "receipt_version": 1,
+  "receipt_id": "replace-with-operator-receipt-id",
+  "created_at": "YYYY-MM-DDTHH:MM:SSZ",
+  "operator": "replace-with-operator-id",
+  "source_cluster": "railiance01",
+  "source_namespace": "openbao",
+  "source_pod": "openbao-0",
+  "cluster_id": "replace-with-live-openbao-cluster-id",
+  "raft_applied_index": 0,
+  "snapshot_created": false,
+  "source_initialized": false,
+  "source_unsealed": false,
+  "snapshot_sha256": "sha256:0000000000000000000000000000000000000000000000000000000000000000",
+  "snapshot_encrypted": false,
+  "encrypted_snapshot_sha256": "sha256:1111111111111111111111111111111111111111111111111111111111111111",
+  "encrypted_location_ref": "replace-with-approved-off-host-custody-handle",
+  "encrypted_copy_off_host": false,
+  "encryption_verified": false,
+  "hash_verified": false,
+  "no_secret_material_recorded": false
+}

docs/railiance01-coordinated-reboot.md

@@ -38,7 +38,9 @@ the reviewed 1/1 Ready audit-core image. The following are human gates and make
 - approved window id and acknowledgements from all five owners;
 - a named attended driver/abort operator;
 - independent provider-console access, not SSH alone;
-- a current non-secret OpenBao Raft snapshot evidence file;
+- a current non-secret OpenBao Raft snapshot receipt that validates against
+  `docs/openbao-reboot-snapshot-receipt.example.json`, the live cluster id and
+  Raft index, and the default 24-hour freshness bound;
 - explicit availability of two of three Shamir shares through their approved
   out-of-band custody paths.
 
@@ -58,6 +60,9 @@ State Hub.
    verify an on-demand backup before continuing.
 5. Create and escrow a fresh OpenBao Raft snapshot through the approved
    encrypted backup lane. Record only its evidence receipt and verification.
+   Pass that receipt to the preflight with
+   `--openbao-snapshot-evidence /path/to/receipt.json`; a merely existing,
+   stale, placeholder, wrong-cluster, unencrypted, or on-host receipt is denied.
 6. Have both Shamir share custodians present. Prove provider-console access and
    SSH access. Record the exact restart and abort operators.
 7. Re-run the preflight with the approval and attestation flags. The result

scripts/audit-core-database-lease-recovery.py

@@ -0,0 +1,438 @@
+#!/usr/bin/env python3
+"""Attended audit-core database lease failure/recovery exercise.
+
+The script is fail-closed and value-safe.  It selects only a single live lease
+under the exact audit-core runtime prefix, never emits its id, never reads a
+database password, never restarts audit-core, and requires a separately
+approved synthetic-load driver.
+"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import importlib.util
+import json
+import os
+import stat
+import subprocess
+import sys
+import time
+from datetime import UTC, datetime, timedelta
+from pathlib import Path
+from typing import Any, Callable
+
+
+ROOT = Path(__file__).resolve().parents[1]
+TASK_ID = "RAILIANCE-WP-0024-T02"
+PROCEDURE = "audit-core-database-lease-recovery"
+LEASE_PREFIX = "database/creds/audit-core-runtime"
+EXTERNAL_SECRET = "audit-core-database"
+CONFIRM = f"{TASK_ID}:attended"
+MAX_WINDOW_SECONDS = 15 * 60
+MIN_LEASE_TTL_SECONDS = 5 * 60
+EXPECTED_DRIVER_KEYS = {
+    "baseline": {"contract_id", "fixture_id", "status", "secret_values_observed"},
+    "expect-unavailable": {
+        "contract_id", "fixture_id", "status", "http_status", "attempts",
+        "secret_values_observed",
+    },
+    "expect-recovered": {
+        "contract_id", "fixture_id", "status", "http_status", "attempts",
+        "secret_values_observed",
+    },
+    "cleanup": {"contract_id", "fixture_id", "status", "secret_values_observed"},
+}
+
+
+class ProcedureError(RuntimeError):
+    pass
+
+
+def load_preflight_module() -> Any:
+    spec = importlib.util.spec_from_file_location(
+        "audit_core_recovery_preflight",
+        ROOT / "scripts" / "audit-core-recovery-preflight.py",
+    )
+    if not spec or not spec.loader:
+        raise ProcedureError("cannot load recovery preflight module")
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
+PREFLIGHT = load_preflight_module()
+
+
+def parse_time(value: str) -> datetime:
+    parsed = datetime.fromisoformat(value.replace("Z", "+00:00"))
+    if parsed.tzinfo is None:
+        raise ProcedureError("approval timestamps must include a timezone")
+    return parsed.astimezone(UTC)
+
+
+def load_approval(path: Path, now: datetime, *, require_open_window: bool) -> dict[str, Any]:
+    try:
+        document = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise ProcedureError("approval receipt is unavailable or invalid JSON") from exc
+    if not isinstance(document, dict):
+        raise ProcedureError("approval receipt must be an object")
+    if document.get("procedure") != PROCEDURE or document.get("task_id") != TASK_ID:
+        raise ProcedureError("approval receipt has the wrong procedure or task")
+    if document.get("status") != "approved" or not document.get("approval_id"):
+        raise ProcedureError("approval receipt is not approved")
+    window = document.get("window") or {}
+    if not window.get("start") or not window.get("end"):
+        raise ProcedureError("approval receipt has no complete window")
+    start, end = parse_time(window["start"]), parse_time(window["end"])
+    if not start < end or (end - start).total_seconds() > MAX_WINDOW_SECONDS:
+        raise ProcedureError("approval window must be positive and at most 15 minutes")
+    if require_open_window and not start <= now <= end:
+        raise ProcedureError("current time is outside the approved window")
+    if not document.get("abort_operator"):
+        raise ProcedureError("approval receipt has no abort operator")
+    owners = document.get("owners") or {}
+    for owner in ("audit-core", "rapp-postgres", "railiance-platform"):
+        record = owners.get(owner) or {}
+        if record.get("acknowledged") is not True or not record.get("message_id"):
+            raise ProcedureError(f"approval receipt lacks {owner} acknowledgement")
+    load = document.get("synthetic_load") or {}
+    if not load.get("contract_id") or not load.get("driver_revision"):
+        raise ProcedureError("approval receipt lacks the synthetic-load contract")
+    return document
+
+
+def safe_run(
+    command: list[str], *, label: str, env: dict[str, str] | None = None
+) -> subprocess.CompletedProcess[str]:
+    completed = subprocess.run(
+        command, text=True, capture_output=True, env=env, check=False
+    )
+    if completed.returncode != 0:
+        # Never attach stdout/stderr: Bao and load-driver processes may hold
+        # sensitive material even though their contract forbids emitting it.
+        raise ProcedureError(f"{label} failed (exit {completed.returncode})")
+    return completed
+
+
+class Authority:
+    def __init__(self, token_file: Path) -> None:
+        if not token_file.is_file() or stat.S_IMODE(token_file.stat().st_mode) != 0o600:
+            raise ProcedureError("OpenBao token file must exist with mode 0600")
+        token = token_file.read_text(encoding="utf-8").splitlines()[0].strip()
+        if not token:
+            raise ProcedureError("OpenBao token file is empty")
+        self.env = dict(os.environ, BAO_ADDR="https://bao.coulomb.social", BAO_TOKEN=token)
+
+    def bao(self, args: list[str], *, label: str) -> str:
+        return safe_run(["bao", *args], label=label, env=self.env).stdout.strip()
+
+    def single_runtime_lease(self) -> dict[str, Any]:
+        raw = self.bao(
+            ["list", "-format=json", f"sys/leases/lookup/{LEASE_PREFIX}"],
+            label="list exact runtime lease handles",
+        )
+        try:
+            handles = json.loads(raw)
+        except json.JSONDecodeError as exc:
+            raise ProcedureError("runtime lease list is invalid JSON") from exc
+        if not isinstance(handles, list) or len(handles) != 1:
+            count = len(handles) if isinstance(handles, list) else "unknown"
+            raise ProcedureError(
+                f"exact runtime prefix must contain one live handle (observed {count})"
+            )
+        lease_id = f"{LEASE_PREFIX}/{handles[0]}"
+        lookup_raw = self.bao(
+            ["write", "-format=json", "sys/leases/lookup", f"lease_id={lease_id}"],
+            label="lookup exact runtime lease metadata",
+        )
+        try:
+            lookup = json.loads(lookup_raw)["data"]
+            issue = parse_time(lookup["issue_time"])
+            expires = parse_time(lookup["expire_time"])
+            ttl = int(lookup["ttl"])
+        except (json.JSONDecodeError, KeyError, TypeError, ValueError) as exc:
+            raise ProcedureError("runtime lease metadata is incomplete") from exc
+        return {
+            "id": lease_id,
+            "fingerprint": hashlib.sha256(lease_id.encode()).hexdigest()[:12],
+            "issue_time": issue,
+            "expire_time": expires,
+            "ttl": ttl,
+        }
+
+    def revoke(self, lease_id: str) -> None:
+        self.bao(["lease", "revoke", lease_id], label="revoke exact runtime lease")
+
+
+class LoadDriver:
+    def __init__(self, path: Path, contract_id: str, expected_revision: str) -> None:
+        if not path.is_file() or not os.access(path, os.X_OK):
+            raise ProcedureError("synthetic-load driver must be an executable file")
+        revision = "sha256:" + hashlib.sha256(path.read_bytes()).hexdigest()
+        if expected_revision != revision:
+            raise ProcedureError("synthetic-load driver does not match the approved revision")
+        self.path = path
+        self.contract_id = contract_id
+        self.revision = revision
+
+    def run(self, phase: str) -> dict[str, Any]:
+        result = safe_run(
+            [str(self.path), phase, "--contract-id", self.contract_id],
+            label=f"synthetic-load {phase}",
+        )
+        try:
+            payload = json.loads(result.stdout)
+        except json.JSONDecodeError as exc:
+            raise ProcedureError(f"synthetic-load {phase} returned invalid JSON") from exc
+        if not isinstance(payload, dict) or set(payload) != EXPECTED_DRIVER_KEYS[phase]:
+            raise ProcedureError(f"synthetic-load {phase} returned an unsafe evidence shape")
+        if payload.get("contract_id") != self.contract_id:
+            raise ProcedureError(f"synthetic-load {phase} returned the wrong contract")
+        if payload.get("secret_values_observed") is not False:
+            raise ProcedureError(f"synthetic-load {phase} did not attest value safety")
+        return payload
+
+
+def secret_state(remote: Any) -> dict[str, Any]:
+    secret_rv = remote.kubectl(
+        ["-n", "audit-core", "get", "secret", EXTERNAL_SECRET, "-o", "jsonpath={.metadata.resourceVersion}"],
+        label="read database Secret resource version",
+    )
+    external = remote.kubectl_json(
+        ["-n", "audit-core", "get", "externalsecret", EXTERNAL_SECRET],
+        label="read database ExternalSecret state",
+    )
+    mount_generation = remote.kubectl(
+        ["-n", "audit-core", "exec", "deploy/audit-core", "--", "readlink", "/etc/audit-core/db/..data"],
+        label="read mounted credential generation",
+    )
+    refresh = external.get("status", {}).get("refreshTime")
+    ready = PREFLIGHT.resource_condition(external)["ready"]
+    if not refresh:
+        raise ProcedureError("database ExternalSecret has no refresh time")
+    return {
+        "resource_version": secret_rv,
+        "mount_generation": mount_generation,
+        "refresh_time": parse_time(refresh),
+        "ready": ready,
+    }
+
+
+def pod_state(remote: Any) -> dict[str, Any]:
+    pods = remote.kubectl_json(
+        ["-n", "audit-core", "get", "pods", "-l", "app.kubernetes.io/name=audit-core"],
+        label="read audit-core pod metadata",
+    ).get("items", [])
+    if len(pods) != 1:
+        raise ProcedureError(f"expected one audit-core pod, observed {len(pods)}")
+    statuses = pods[0].get("status", {}).get("containerStatuses", [])
+    return {
+        "uid": pods[0]["metadata"]["uid"],
+        "restart_count": sum(int(item.get("restartCount", 0)) for item in statuses),
+    }
+
+
+def assert_lease_matches_refresh(lease: dict[str, Any], secret: dict[str, Any]) -> None:
+    delta = abs((lease["issue_time"] - secret["refresh_time"]).total_seconds())
+    if (
+        delta > 5
+        or secret["refresh_time"] < lease["issue_time"] - timedelta(seconds=5)
+        or secret["refresh_time"] > lease["expire_time"]
+    ):
+        raise ProcedureError("single runtime lease is not coherent with the mounted Secret refresh")
+    if lease["ttl"] < MIN_LEASE_TTL_SECONDS:
+        raise ProcedureError("runtime lease is too close to automatic refresh; wait for the next sync")
+
+
+def wait_for(
+    predicate: Callable[[], bool], *, label: str, timeout: int = 120, interval: float = 2
+) -> None:
+    deadline = time.monotonic() + timeout
+    while time.monotonic() < deadline:
+        if predicate():
+            return
+        time.sleep(interval)
+    raise ProcedureError(f"timed out waiting for {label}")
+
+
+def force_refresh(remote: Any, baseline: dict[str, Any]) -> dict[str, Any]:
+    remote.kubectl(
+        [
+            "-n", "audit-core", "annotate", "externalsecret", EXTERNAL_SECRET,
+            f"railiance.io/force-sync={int(time.time())}", "--overwrite",
+        ],
+        label="force database ExternalSecret reconciliation",
+    )
+    latest: dict[str, Any] = {}
+
+    def changed() -> bool:
+        nonlocal latest
+        latest = secret_state(remote)
+        return bool(
+            latest["ready"]
+            and latest["resource_version"] != baseline["resource_version"]
+            and latest["mount_generation"] != baseline["mount_generation"]
+            and latest["refresh_time"] > baseline["refresh_time"]
+        )
+
+    wait_for(changed, label="database Secret and mounted generation refresh")
+    return latest
+
+
+def exercise(args: argparse.Namespace) -> dict[str, Any]:
+    approval = load_approval(args.approval, datetime.now(UTC), require_open_window=True)
+    if args.confirm != CONFIRM:
+        raise ProcedureError(f"live exercise requires --confirm {CONFIRM}")
+    remote = PREFLIGHT.Remote(args.remote)
+    authority = Authority(args.token_file)
+    driver = LoadDriver(
+        args.load_driver,
+        approval["synthetic_load"]["contract_id"],
+        approval["synthetic_load"]["driver_revision"],
+    )
+
+    preflight_args = argparse.Namespace(
+        approved_window_id=approval["approval_id"],
+        audit_core_owner_ack=True,
+        rapp_postgres_owner_ack=True,
+        synthetic_load_id=approval["synthetic_load"]["contract_id"],
+        abort_operator=approval["abort_operator"],
+    )
+    preflight = PREFLIGHT.database_lease_preflight(remote, preflight_args)
+    if not preflight["ready_for_live_execution"]:
+        raise ProcedureError("database lease recovery preflight is not ready")
+
+    before_pod = pod_state(remote)
+    before_secret = secret_state(remote)
+    lease = authority.single_runtime_lease()
+    assert_lease_matches_refresh(lease, before_secret)
+    baseline_load = driver.run("baseline")
+    if baseline_load.get("status") != "ready":
+        raise ProcedureError("synthetic-load baseline is not ready")
+    revoked = False
+    recovered_secret: dict[str, Any] | None = None
+    unavailable: dict[str, Any] | None = None
+    recovered: dict[str, Any] | None = None
+    cleanup_result: dict[str, Any] | None = None
+    completed = False
+    try:
+        # Close the race with ESO's ordinary refresh before the destructive step.
+        if secret_state(remote) != before_secret or pod_state(remote) != before_pod:
+            raise ProcedureError("baseline changed before revocation")
+        current = authority.single_runtime_lease()
+        if current["fingerprint"] != lease["fingerprint"]:
+            raise ProcedureError("runtime lease changed before revocation")
+        authority.revoke(lease["id"])
+        revoked = True
+        wait_for(
+            lambda: PREFLIGHT.endpoint_status(remote, "/healthz") == 200
+            and PREFLIGHT.endpoint_status(remote, "/readyz") == 503,
+            label="health 200 and readiness 503 after revocation",
+            timeout=60,
+        )
+        unavailable = driver.run("expect-unavailable")
+        if (
+            unavailable.get("http_status") != 503
+            or unavailable.get("status") != "retryable_unavailable"
+            or not isinstance(unavailable.get("attempts"), int)
+            or unavailable["attempts"] < 1
+            or unavailable.get("fixture_id") != baseline_load.get("fixture_id")
+        ):
+            raise ProcedureError("synthetic load did not prove retryable 503")
+        recovered_secret = force_refresh(remote, before_secret)
+        wait_for(
+            lambda: PREFLIGHT.endpoint_status(remote, "/readyz") == 200,
+            label="audit-core readiness recovery",
+        )
+        recovered = driver.run("expect-recovered")
+        if (
+            recovered.get("status") not in {"accepted", "duplicate"}
+            or recovered.get("http_status") not in {200, 202}
+            or recovered.get("fixture_id") != baseline_load.get("fixture_id")
+        ):
+            raise ProcedureError("synthetic load did not prove accepted/duplicate recovery")
+        after_pod = pod_state(remote)
+        if after_pod != before_pod:
+            raise ProcedureError("audit-core pod identity or restart count changed")
+        completed = True
+    finally:
+        if revoked and recovered_secret is None:
+            try:
+                force_refresh(remote, before_secret)
+            except Exception:
+                pass
+        try:
+            cleanup_result = driver.run("cleanup")
+            if (
+                cleanup_result.get("status") != "clean"
+                or cleanup_result.get("fixture_id") != baseline_load.get("fixture_id")
+            ):
+                raise ProcedureError("synthetic-load cleanup did not confirm exact fixture removal")
+        except Exception:
+            if completed:
+                raise
+    return {
+        "procedure": PROCEDURE,
+        "task_id": TASK_ID,
+        "approval_id": approval["approval_id"],
+        "synthetic_load_contract": approval["synthetic_load"]["contract_id"],
+        "synthetic_load_driver_revision": driver.revision,
+        "lease_handle_fingerprint": lease["fingerprint"],
+        "lease_revoked": revoked,
+        "health_during_failure": 200,
+        "readiness_during_failure": 503,
+        "synthetic_unavailable_attempts": unavailable["attempts"] if unavailable else None,
+        "secret_resource_version_changed": bool(
+            recovered_secret
+            and recovered_secret["resource_version"] != before_secret["resource_version"]
+        ),
+        "mount_generation_changed": bool(
+            recovered_secret
+            and recovered_secret["mount_generation"] != before_secret["mount_generation"]
+        ),
+        "readiness_recovered": PREFLIGHT.endpoint_status(remote, "/readyz") == 200,
+        "synthetic_recovery_status": recovered["status"] if recovered else None,
+        "same_pod_uid": pod_state(remote)["uid"] == before_pod["uid"],
+        "restart_count_unchanged": pod_state(remote)["restart_count"] == before_pod["restart_count"],
+        "load_cleanup_status": cleanup_result["status"] if cleanup_result else None,
+        "completed_at": datetime.now(UTC).replace(microsecond=0).isoformat().replace("+00:00", "Z"),
+        "secret_values_observed": False,
+    }
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("command", choices=["validate-approval", "exercise"])
+    parser.add_argument("--approval", type=Path, required=True)
+    parser.add_argument("--remote", default="railiance01")
+    parser.add_argument(
+        "--token-file", type=Path,
+        default=Path.home() / ".local/openbao/platform-admin.token",
+    )
+    parser.add_argument("--load-driver", type=Path)
+    parser.add_argument("--confirm")
+    args = parser.parse_args()
+    try:
+        if args.command == "validate-approval":
+            approval = load_approval(args.approval, datetime.now(UTC), require_open_window=False)
+            result = {
+                "procedure": PROCEDURE,
+                "approval_id": approval["approval_id"],
+                "approval_receipt_valid": True,
+                "secret_values_observed": False,
+            }
+        else:
+            if args.load_driver is None:
+                raise ProcedureError("live exercise requires --load-driver")
+            result = exercise(args)
+    except (OSError, IndexError, ProcedureError, PREFLIGHT.PreflightError) as exc:
+        print(f"database lease recovery failed: {exc}", file=sys.stderr)
+        return 1
+    print(json.dumps(result, indent=2, sort_keys=True))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

scripts/audit-core-recovery-preflight.py

@@ -111,6 +111,80 @@ def secret_key_names(description: str) -> set[str]:
     return keys
 
 
+def validate_snapshot_receipt(
+    path: Path,
+    *,
+    live_openbao: dict[str, Any],
+    now: datetime,
+    max_age_hours: float,
+) -> dict[str, Any]:
+    try:
+        receipt = json.loads(path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise PreflightError("OpenBao snapshot receipt is unavailable or invalid JSON") from exc
+    if not isinstance(receipt, dict) or receipt.get("receipt_version") != 1:
+        raise PreflightError("OpenBao snapshot receipt has an unsupported shape/version")
+    required_strings = (
+        "receipt_id", "created_at", "operator", "source_cluster",
+        "source_namespace", "source_pod", "cluster_id", "snapshot_sha256",
+        "encrypted_snapshot_sha256", "encrypted_location_ref",
+    )
+    for key in required_strings:
+        if not isinstance(receipt.get(key), str) or not receipt[key].strip():
+            raise PreflightError(f"OpenBao snapshot receipt is missing {key}")
+    required_true = (
+        "snapshot_created", "source_initialized", "source_unsealed",
+        "snapshot_encrypted", "encrypted_copy_off_host", "encryption_verified",
+        "hash_verified", "no_secret_material_recorded",
+    )
+    for key in required_true:
+        if receipt.get(key) is not True:
+            raise PreflightError(f"OpenBao snapshot receipt requires {key}=true")
+    if (
+        receipt["source_cluster"] != "railiance01"
+        or receipt["source_namespace"] != "openbao"
+        or receipt["source_pod"] != "openbao-0"
+        or receipt["cluster_id"] != live_openbao.get("cluster_id")
+    ):
+        raise PreflightError("OpenBao snapshot receipt does not match the live source")
+    try:
+        created = rfc3339(receipt["created_at"])
+        snapshot_index = int(receipt["raft_applied_index"])
+        live_index = int(live_openbao["raft_applied_index"])
+    except (KeyError, TypeError, ValueError) as exc:
+        raise PreflightError("OpenBao snapshot receipt has invalid time/index metadata") from exc
+    age_hours = (now - created).total_seconds() / 3600
+    if age_hours < -(5 / 60) or age_hours > max_age_hours:
+        raise PreflightError("OpenBao snapshot receipt is outside the permitted age")
+    if snapshot_index <= 0 or snapshot_index > live_index:
+        raise PreflightError("OpenBao snapshot receipt has an impossible Raft index")
+    digest_pattern = re.compile(r"^sha256:[0-9a-f]{64}$")
+    digests = (receipt["snapshot_sha256"], receipt["encrypted_snapshot_sha256"])
+    if any(not digest_pattern.fullmatch(value) for value in digests):
+        raise PreflightError("OpenBao snapshot receipt has an invalid digest")
+    if any(len(set(value.removeprefix("sha256:"))) <= 1 for value in digests):
+        raise PreflightError("OpenBao snapshot receipt contains a placeholder digest")
+    if digests[0] == digests[1]:
+        raise PreflightError("plain and encrypted snapshot digests must differ")
+    encoded = json.dumps(receipt, sort_keys=True)
+    for marker in (
+        "BEGIN PRIVATE KEY", "BEGIN OPENSSH PRIVATE KEY", "AGE-SECRET-KEY-1",
+        "OPENBAO_ROOT_TOKEN", "VAULT_TOKEN", "hvs.", "<", "YYYY-MM-DD",
+    ):
+        if marker in encoded:
+            raise PreflightError("OpenBao snapshot receipt contains forbidden material/placeholder")
+    return {
+        "receipt_id": receipt["receipt_id"],
+        "created_at": created.isoformat().replace("+00:00", "Z"),
+        "age_hours": round(age_hours, 2),
+        "cluster_id_matches": True,
+        "raft_applied_index": snapshot_index,
+        "encrypted_copy_off_host": True,
+        "verified": True,
+        "secret_values_observed": False,
+    }
+
+
 def common_state(remote: Remote, now: datetime) -> dict[str, Any]:
     node_list = remote.kubectl_json(["get", "nodes"], label="read node state")
     nodes = node_list.get("items", [])
@@ -186,6 +260,8 @@ def common_state(remote: Remote, now: datetime) -> dict[str, Any]:
             "seal_type": bao_status.get("type"),
             "threshold": bao_status.get("t"),
             "shares": bao_status.get("n"),
+            "cluster_id": bao_status.get("cluster_id"),
+            "raft_applied_index": bao_status.get("raft_applied_index"),
         },
         "external_secret_stores": stores,
         "external_secrets": external_secrets,
@@ -220,6 +296,23 @@ def automated_common_pass(state: dict[str, Any]) -> bool:
     )
 
 
+def endpoint_status(remote: Remote, path: str) -> int:
+    code = remote.kubectl(
+        [
+            "-n", "audit-core", "exec", "deploy/audit-core", "--", "python", "-c",
+            "import urllib.request,urllib.error; "
+            f"u='http://127.0.0.1:8080{path}'; "
+            "\ntry:\n r=urllib.request.urlopen(u,timeout=3); print(r.status)"
+            "\nexcept urllib.error.HTTPError as e:\n print(e.code)",
+        ],
+        label=f"probe audit-core {path}",
+    )
+    try:
+        return int(code)
+    except ValueError as exc:
+        raise PreflightError(f"audit-core {path} returned an invalid status") from exc
+
+
 def database_lease_preflight(remote: Remote, args: argparse.Namespace) -> dict[str, Any]:
     now = datetime.now(UTC)
     state = common_state(remote, now)
@@ -232,14 +325,8 @@ def database_lease_preflight(remote: Remote, args: argparse.Namespace) -> dict[s
         label="read database Secret key names",
     )
     keys = secret_key_names(description)
-    health_code = remote.kubectl(
+    health_code = endpoint_status(remote, "/healthz")
-        ["-n", "audit-core", "exec", "deploy/audit-core", "--", "python", "-c", 'import urllib.request; print(urllib.request.urlopen("http://127.0.0.1:8080/healthz", timeout=3).status)'],
+    ready_code = endpoint_status(remote, "/readyz")
-        label="probe audit-core health",
-    )
-    ready_code = remote.kubectl(
-        ["-n", "audit-core", "exec", "deploy/audit-core", "--", "python", "-c", 'import urllib.request; print(urllib.request.urlopen("http://127.0.0.1:8080/readyz", timeout=3).status)'],
-        label="probe audit-core readiness",
-    )
     gates = {
         "approved_window_id_recorded": bool(args.approved_window_id),
         "audit_core_owner_acknowledged": args.audit_core_owner_ack,
@@ -247,7 +334,7 @@ def database_lease_preflight(remote: Remote, args: argparse.Namespace) -> dict[s
         "approved_synthetic_load_contract": bool(args.synthetic_load_id),
         "attended_abort_operator_named": bool(args.abort_operator),
     }
-    automated = automated_common_pass(state) and keys == EXPECTED_DB_KEYS and health_code == "200" and ready_code == "200"
+    automated = automated_common_pass(state) and keys == EXPECTED_DB_KEYS and health_code == 200 and ready_code == 200
     return {
         "procedure": "audit-core-database-lease-recovery",
         "preflight_only": True,
@@ -257,8 +344,8 @@ def database_lease_preflight(remote: Remote, args: argparse.Namespace) -> dict[s
         "baseline": {
             "database_secret_resource_version": secret_rv,
             "database_secret_keys": sorted(keys),
-            "health_status": int(health_code),
+            "health_status": health_code,
-            "readiness_status": int(ready_code),
+            "readiness_status": ready_code,
             "pod_uid": state["audit_core"]["pod_uid"],
             "restart_count": state["audit_core"]["restart_count"],
         },
@@ -276,9 +363,17 @@ def reboot_preflight(remote: Remote, args: argparse.Namespace) -> dict[str, Any]
     )
     latest = latest_completed_backup(backups, now)
     snapshot_evidence = Path(args.openbao_snapshot_evidence).resolve() if args.openbao_snapshot_evidence else None
+    snapshot_receipt = None
+    if snapshot_evidence:
+        snapshot_receipt = validate_snapshot_receipt(
+            snapshot_evidence,
+            live_openbao=state["openbao"],
+            now=now,
+            max_age_hours=args.max_snapshot_age_hours,
+        )
     gates = {
         "approved_window_id_recorded": bool(args.approved_window_id),
-        "openbao_snapshot_evidence_present": bool(snapshot_evidence and snapshot_evidence.is_file()),
+        "openbao_snapshot_evidence_present": bool(snapshot_receipt),
         "unseal_quorum_attested": args.unseal_quorum_attested,
         "provider_console_access_attested": args.provider_console_attested,
         "host_cluster_platform_database_audit_owners_acknowledged": args.all_owners_ack,
@@ -293,6 +388,8 @@ def reboot_preflight(remote: Remote, args: argparse.Namespace) -> dict[str, Any]
         "operator_gates": gates,
         "latest_platform_pg_backup": latest,
         "max_backup_age_hours": args.max_backup_age_hours,
+        "openbao_snapshot_receipt": snapshot_receipt,
+        "max_snapshot_age_hours": args.max_snapshot_age_hours,
         "state": state,
         "secret_values_observed": False,
     }
@@ -312,6 +409,7 @@ def main() -> int:
     parser.add_argument("--provider-console-attested", action="store_true")
     parser.add_argument("--all-owners-ack", action="store_true")
     parser.add_argument("--max-backup-age-hours", type=float, default=26.0)
+    parser.add_argument("--max-snapshot-age-hours", type=float, default=24.0)
     args = parser.parse_args()
     try:
         remote = Remote(args.remote)

tests/test_audit_core_database_lease_recovery.py

@@ -0,0 +1,87 @@
+from __future__ import annotations
+
+import importlib.util
+import json
+import os
+import tempfile
+import unittest
+from datetime import UTC, datetime, timedelta
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+SPEC = importlib.util.spec_from_file_location(
+    "audit_core_database_lease_recovery",
+    ROOT / "scripts" / "audit-core-database-lease-recovery.py",
+)
+assert SPEC and SPEC.loader
+module = importlib.util.module_from_spec(SPEC)
+SPEC.loader.exec_module(module)
+
+
+def approved_receipt(now: datetime) -> dict:
+    return {
+        "procedure": module.PROCEDURE,
+        "task_id": module.TASK_ID,
+        "status": "approved",
+        "approval_id": "operator-window-1",
+        "window": {
+            "start": (now - timedelta(minutes=1)).isoformat(),
+            "end": (now + timedelta(minutes=9)).isoformat(),
+        },
+        "abort_operator": "operator-a",
+        "owners": {
+            owner: {"acknowledged": True, "message_id": f"ack-{owner}"}
+            for owner in ("audit-core", "rapp-postgres", "railiance-platform")
+        },
+        "synthetic_load": {"contract_id": "load-1", "driver_revision": "abc123"},
+    }
+
+
+class DatabaseLeaseRecoveryTests(unittest.TestCase):
+    def test_approval_requires_exact_owner_and_window_contract(self) -> None:
+        now = datetime(2026, 8, 22, 12, 0, tzinfo=UTC)
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "approval.json"
+            path.write_text(json.dumps(approved_receipt(now)), encoding="utf-8")
+            result = module.load_approval(path, now, require_open_window=True)
+        self.assertEqual("operator-window-1", result["approval_id"])
+
+    def test_pending_or_overlong_approval_is_rejected(self) -> None:
+        now = datetime(2026, 8, 22, 12, 0, tzinfo=UTC)
+        receipt = approved_receipt(now)
+        receipt["status"] = "pending"
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "approval.json"
+            path.write_text(json.dumps(receipt), encoding="utf-8")
+            with self.assertRaises(module.ProcedureError):
+                module.load_approval(path, now, require_open_window=True)
+
+    def test_lease_must_align_to_refresh_and_have_headroom(self) -> None:
+        issue = datetime(2026, 8, 22, 12, 0, tzinfo=UTC)
+        lease = {
+            "issue_time": issue,
+            "expire_time": issue + timedelta(minutes=15),
+            "ttl": 800,
+        }
+        secret = {"refresh_time": issue + timedelta(seconds=1)}
+        module.assert_lease_matches_refresh(lease, secret)
+        lease["ttl"] = 20
+        with self.assertRaises(module.ProcedureError):
+            module.assert_lease_matches_refresh(lease, secret)
+
+    def test_load_driver_rejects_extra_evidence_fields(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            driver = Path(directory) / "driver"
+            driver.write_text(
+                "#!/bin/sh\nprintf '%s\\n' '{\"contract_id\":\"load-1\",\"fixture_id\":\"f-1\",\"status\":\"ready\",\"secret_values_observed\":false,\"token\":\"forbidden\"}'\n",
+                encoding="utf-8",
+            )
+            os.chmod(driver, 0o700)
+            revision = "sha256:" + module.hashlib.sha256(driver.read_bytes()).hexdigest()
+            with self.assertRaises(module.ProcedureError):
+                module.LoadDriver(driver, "load-1", revision).run("baseline")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_audit_core_recovery_preflight.py

@@ -1,8 +1,10 @@
 from __future__ import annotations
 
 import importlib.util
+import json
+import tempfile
 import unittest
-from datetime import UTC, datetime
+from datetime import UTC, datetime, timedelta
 from pathlib import Path
 
 
@@ -96,6 +98,78 @@ host:      50 bytes
             {"username", "password", "host"}, module.secret_key_names(description)
         )
 
+    def test_snapshot_receipt_matches_live_cluster_and_freshness(self) -> None:
+        now = datetime(2026, 8, 22, 12, 0, tzinfo=UTC)
+        receipt = {
+            "receipt_version": 1,
+            "receipt_id": "snapshot-20260822",
+            "created_at": (now - timedelta(hours=1)).isoformat(),
+            "operator": "operator-a",
+            "source_cluster": "railiance01",
+            "source_namespace": "openbao",
+            "source_pod": "openbao-0",
+            "cluster_id": "cluster-1",
+            "raft_applied_index": 900,
+            "snapshot_created": True,
+            "source_initialized": True,
+            "source_unsealed": True,
+            "snapshot_sha256": "sha256:" + "ab" * 32,
+            "snapshot_encrypted": True,
+            "encrypted_snapshot_sha256": "sha256:" + "cd" * 32,
+            "encrypted_location_ref": "custody:openbao:snapshot-20260822",
+            "encrypted_copy_off_host": True,
+            "encryption_verified": True,
+            "hash_verified": True,
+            "no_secret_material_recorded": True,
+        }
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "receipt.json"
+            path.write_text(json.dumps(receipt), encoding="utf-8")
+            result = module.validate_snapshot_receipt(
+                path,
+                live_openbao={"cluster_id": "cluster-1", "raft_applied_index": 1000},
+                now=now,
+                max_age_hours=24,
+            )
+        self.assertTrue(result["verified"])
+        self.assertEqual(1.0, result["age_hours"])
+        self.assertFalse(result["secret_values_observed"])
+
+    def test_snapshot_receipt_rejects_stale_or_secret_material(self) -> None:
+        now = datetime(2026, 8, 22, 12, 0, tzinfo=UTC)
+        receipt = {
+            "receipt_version": 1,
+            "receipt_id": "snapshot-old",
+            "created_at": (now - timedelta(hours=25)).isoformat(),
+            "operator": "operator-a",
+            "source_cluster": "railiance01",
+            "source_namespace": "openbao",
+            "source_pod": "openbao-0",
+            "cluster_id": "cluster-1",
+            "raft_applied_index": 900,
+            "snapshot_created": True,
+            "source_initialized": True,
+            "source_unsealed": True,
+            "snapshot_sha256": "sha256:" + "ab" * 32,
+            "snapshot_encrypted": True,
+            "encrypted_snapshot_sha256": "sha256:" + "cd" * 32,
+            "encrypted_location_ref": "custody:old",
+            "encrypted_copy_off_host": True,
+            "encryption_verified": True,
+            "hash_verified": True,
+            "no_secret_material_recorded": True,
+        }
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "receipt.json"
+            path.write_text(json.dumps(receipt), encoding="utf-8")
+            with self.assertRaises(module.PreflightError):
+                module.validate_snapshot_receipt(
+                    path,
+                    live_openbao={"cluster_id": "cluster-1", "raft_applied_index": 1000},
+                    now=now,
+                    max_age_hours=24,
+                )
+
 
 if __name__ == "__main__":
     unittest.main()

workplans/RAILIANCE-WP-0024-audit-core-custody-and-recovery-coordination.md

@@ -113,6 +113,20 @@ checks without observing Secret data. The live revocation still waits for an
 approved window, audit-core/rapp-postgres acknowledgement, a separately
 approved synthetic-load contract, and a named abort operator.
 
+**Harness hardening (2026-08-22):**
+`scripts/audit-core-database-lease-recovery.py` now implements the attended
+orchestration but cannot run from repository access alone. It requires a
+value-safe approved receipt based on
+`docs/audit-core-database-lease-approval.example.json`, a revision-pinned
+synthetic-load driver, exact confirmation, an open ≤15-minute window, all
+three owner acknowledgements, and at least five minutes before the ordinary
+lease refresh. OpenBao metadata inspection corrected the selection contract:
+lease lookup exposes issue/expiry/TTL, not a database username. The harness
+therefore aborts unless the exact runtime prefix has one live handle whose
+issue time coheres with the ExternalSecret refresh. A live read-only check
+observed exactly one handle, 0.073 seconds of issue/refresh skew, and no secret
+values. It was not revoked.
+
 ## T03 — Define the coordinated railiance01 reboot exercise
 
 ```task
@@ -144,6 +158,16 @@ the intentionally non-automated gates: an approved window and owner set,
 provider-console access, a current OpenBao Raft snapshot receipt, 2-of-3 Shamir
 quorum availability, and a named abort operator.
 
+**Snapshot gate hardening (2026-08-22):** reboot preflight no longer accepts a
+file-existence assertion as snapshot evidence. A receipt based on
+`docs/openbao-reboot-snapshot-receipt.example.json` must be no more than 24
+hours old, match the live OpenBao cluster id and a possible Raft applied index,
+carry distinct non-placeholder plain/encrypted hashes, and attest encryption,
+off-host persistence, verification, and absence of secret material. The
+pending example intentionally fails. Creating the secret-bearing snapshot and
+its off-host copy remains an attended platform-custody action near the future
+reboot window; no snapshot was created in this implementation step.
+
 ## T04 — Route the contracts and approval gates
 
 ```task