diff --git a/docs/architecture/rfcs/monorepo-distribution-split-v0.md b/docs/architecture/rfcs/monorepo-distribution-split-v0.md
index e7dcd45329..f6c6dafbcf 100644
--- a/docs/architecture/rfcs/monorepo-distribution-split-v0.md
+++ b/docs/architecture/rfcs/monorepo-distribution-split-v0.md
@@ -1619,6 +1619,61 @@ invalid Vision/workspace also remain refusals. The temporary measurement harness
 and raw captures stay private; this recipe describes the repeated workload and
 its oracles, not a distributed benchmark runner or public reproduction receipt.
 
+### 2026-11-07 — Original-Turn closeout survives automatic Todo archival
+
+The complete CLI comparison in #8257 exposed a lifecycle failure with 101
+synthetic Todos. Fresh source-console reproductions on `4a33eb4b5` retain
+`todo_archived` on both File and default SQLite after ordinary completion,
+durable writeback and one quota debit. The frozen comparison remains unchanged;
+these failed episodes are not accepted latency samples.
+
+The proposed repair stays in `todo_terminal_lifecycle.ts`. Its existing
+same-Turn fence and exact ordinary-completion receipt verification also qualify
+an archived Todo's closeout before the active-only rejection. One local helper
+owns both active and archived parent checks. Original claim/lease, task digest,
+validation declaration/receipt and provider CAS checks remain in force; archive
+selection and retention thresholds are unchanged. This permits only the
+original completion's terminal phase, without reopening archived work.
+
+Real File/SQLite console fixtures now complete the large-state sequence and
+replay completion, writeback, quota spend and terminal closeout with one debit
+and an unchanged independently verified JSON result. The durable CLI regression
+uses a real validator and ordinary hard-lease claim: premature settlement,
+unrelated archived work, another actor, claim or Turn, and changed completion
+intent are refused. Terminal closeout retains the original passed validation
+receipt, first completion time and archive state without rerunning the validator.
+Existing provider conformance retains source/digest, validation, lease and replay
+checks; an isolated real PostgreSQL suite qualifies its shared owner separately.
+
+The author proposal also integrates main `abf144202` (#8161), whose explicit
+Goal-scoped checks do not gate local Todo completion. The archive regression
+covers a failed Goal commitment alongside the local validator: ordinary and
+terminal completion still preserve the original receipt and one debit, never
+rerun the Goal checker, and leave the overall check failed. This is author-side
+integration coverage; it does not replace independent review or reviewed-repair
+wheel qualification.
+
+Independent review of #8264 at `12627eff8` found that changing only `--evidence`
+could replace the archived Todo's accepted artifact while returning the old
+validation receipt. All four File/SQLite × Goal-check fixtures reproduce this
+failure on that head. The revision integrates target main `10f66ec5e` and rejects
+different supplied evidence before archived closeout. Its new archived-closeout
+receipt also retains the accepted evidence, so a changed retry cannot bypass the
+check through receipt recovery. Omitted or empty evidence with a valid completion
+note keeps the original value; the CLI's completion-note requirement still applies.
+Historical fingerprints and older receipts remain compatible; active
+completion behavior is unchanged. The regression checks canonical evidence,
+original validation and completion time, and once-only validation/writeback/debit
+before and after terminal replay. The earlier reviewed head remains unaccepted;
+the revised head still needs independent review.
+
+This is a proposed lifecycle repair enabling the remaining R1/V4 journey and
+large-state comparison. Independent exact-head review, reviewed-repair wheel
+readback and remaining guard/parser characterization still govern the next
+step. It does not upgrade an installed host or qualify interruption/restart,
+split installation, full R0/V10/R1, M1/#8011, D2/D5/D6/D3, context/model,
+adoption or complete-cost benefit. I3 remains at least one minor release.
+
 ## Appendix B: Decision log
 
 | Date | Decision | Owner / approval | Alternatives | Normative sections changed |
diff --git a/docs/architecture/rfcs/monorepo-distribution-split-v0.zh-CN.md b/docs/architecture/rfcs/monorepo-distribution-split-v0.zh-CN.md
index d1ae12fed5..f1ac16c0ce 100644
--- a/docs/architecture/rfcs/monorepo-distribution-split-v0.zh-CN.md
+++ b/docs/architecture/rfcs/monorepo-distribution-split-v0.zh-CN.md
@@ -1223,6 +1223,51 @@ stale claim 使用普通 release 后的 lease。它们必须在 writeback/debit
 Vision/workspace 同样保留拒绝。临时测量 harness 和 raw capture 保持私有；这个 recipe
 描述重复的 workload/oracle，不是分发的 benchmark runner 或公共复现 receipt。
 
+### 2026-11-07 — 原 Turn 终结在 Todo 自动归档后仍可完成
+
+#8257 的完整 CLI 比较在 101 个合成 Todo 下暴露了生命周期故障。基于
+`4a33eb4b5` 的新源码 console 复现在 File 和默认 SQLite 上均保留同一
+`todo_archived`：普通完成、持久写回和一次额度扣账已提交。冻结比较保持原样，
+这些失败 episode 不计入验收通过的延迟样本。
+
+修复提案仍由 `todo_terminal_lifecycle.ts` 拥有。在仅限 active 的拒绝之前，
+复用已有同 Turn fence 与精确普通 completion receipt 校验，验证已归档 Todo
+的终结资格。一个文件内 helper 统一 active 和已归档路径的父 receipt 检查。
+原 claim/lease、任务 digest、validation 声明与 receipt、provider CAS 校验
+继续生效；归档选择与保留阈值保持原样。允许的只是原 completion 的终结阶段，
+不会重新开放已归档工作。
+
+真实 File/SQLite console fixture 现在可以完成大状态整轮，并回放 completion、
+写回、扣账和 terminal closeout：只有一次扣账，独立验证的 JSON 结果保持原样。
+持久 CLI 回归使用真实 validator 和普通 hard-lease claim，拒绝过早结算、
+无关归档工作、其他 actor、claim 或 Turn，以及变更后的 completion intent。
+终结保留原已通过的 validation receipt、首次完成时间和归档状态，不重复执行
+validator。既有 provider conformance 继续检查 source/digest、validation、
+lease 和 replay；隔离真实 PostgreSQL suite 单独验证其共享 owner。
+
+作者提案还集成了 main `abf144202`（#8161）：显式 Goal 范围的检查不约束局部
+Todo 完成。归档回归在局部 validator 之外加入未通过的整体 Goal 承诺检查：
+普通完成和终结仍保留原 receipt 与一次扣账，不重复执行 Goal checker，整体
+检查继续失败。这是作者阶段的集成覆盖，不能替代独立审阅或审阅修复后的
+wheel 资格验证。
+
+#8264 `12627eff8` 的独立审阅发现：只改 `--evidence` 就能替换已归档 Todo
+接受的产物，却仍返回旧 validation receipt。File/SQLite × 有无整体 Goal 检查
+的四个 fixture 均在该 head 复现失败。修订集成 target main `10f66ec5e`，在归档
+终结前拒绝不同的显式 evidence；新归档 closeout receipt 同时保存已接受证据，
+让变更后的重试不能通过 receipt recovery 绕过检查。带有效 completion note 时，
+省略或空 evidence 保留原值；CLI 原有的完成说明要求继续适用。
+历史 fingerprint 和旧 receipt 继续兼容，active completion 行为保持原样。
+回归在 terminal replay 前后检查 canonical evidence、原 validation 和完成时间，
+以及只执行一次的 validation/writeback/debit。此前被审阅的 head 仍未获验收，
+修订 head 仍需独立审阅。
+
+这是支持后续 R1/V4 journey 和大状态比较的生命周期修复提案。下一步仍需
+独立 exact-head review、审阅修复后的 wheel 回读，以及剩余 guard/parser
+刻画。它不会升级 installed host，也不验收中断/重启、split installation、
+完整 R0/V10/R1、M1/#8011、D2/D5/D6/D3、context/model、adoption 或完整成本
+收益。I3 继续保留至少一个 minor release。
+
 ## 附录 B：决策日志
 
 | 日期 | 决策 | 负责人 / 批准 | 备选 | 变更的规范章节 |
diff --git a/loopx/control_plane/coordination/todo_terminal_lifecycle.ts b/loopx/control_plane/coordination/todo_terminal_lifecycle.ts
index c7c9fb11da..387f7e65f1 100644
--- a/loopx/control_plane/coordination/todo_terminal_lifecycle.ts
+++ b/loopx/control_plane/coordination/todo_terminal_lifecycle.ts
@@ -487,6 +487,15 @@ function terminalFailure(
   };
 }
 
+function completionEvidenceRejection(
+  input: CoordinationTodoTerminalLifecycleInput,
+  acceptedEvidence: unknown,
+): CoordinationTodoTerminalLifecycleResult | null {
+  if (input.evidence === null || input.evidence === (acceptedEvidence ?? null)) return null;
+  return terminalFailure("terminal_completion_receipt_mismatch",
+    "Archived closeout must retain the evidence accepted by ordinary completion.", {}, "decision_rejection");
+}
+
 function terminalRequestSha(input: CoordinationTodoTerminalLifecycleInput): string {
   const completionPolicyIdentity = input.completion_policy_request === null
     ? null
@@ -1194,6 +1203,59 @@ function semanticDuplicate(
   return undefined;
 }
 
+async function readSameTurnCompletionReceipt(
+  store: AuthorityStore,
+  input: ResolvedCoordinationTodoTerminalLifecycleInput,
+  normalized: CoordinationTodoTerminalLifecycleInput,
+  todo: JsonObject,
+): Promise<CoordinationTodoTerminalLifecycleResult | null> {
+  if (input.command !== "complete" || normalized.operation_identity.kind !== "completion_turn" ||
+      !input.requested_no_followup) return null;
+  const validationRequired = todo.completion_validation_required === true;
+  const validationSha256 = todo.completion_validation_sha256;
+  let fence;
+  try {
+    fence = evaluateTodoCompletionFence({schema_version: TODO_COMPLETION_FENCE_REQUEST_SCHEMA,
+      projection_source: "materialized", goal_id: input.goal_id, todo_id: input.todo_id, todo,
+      requested_no_followup: input.requested_no_followup,
+      requested_completion_turn_key: input.requested_completion_turn_key,
+      requested_completion_identity_source: input.requested_completion_identity_source});
+  } catch (error) {
+    return terminalFailure("invalid_todo_completion_transaction",
+      error instanceof Error ? error.message : "Invalid original completion identity");
+  }
+  if (fence.outcome === "continue" && fence.reason === "same_turn_terminal_upgrade") {
+    const originalInput = {...input,
+      ...(normalized.claim_operation_id === undefined ? {} : {claim_operation_id: normalized.claim_operation_id,
+        lease_idempotency_key: normalized.lease_idempotency_key, lease_expected_version: normalized.lease_expected_version}),
+      requested_no_followup: false,
+      operation_id: completionTurnOperationId(input, false)};
+    const original = await terminalReceipt(originalInput, terminalRequestSha(originalInput)).read(store);
+    if (original === null) return terminalFailure("terminal_completion_receipt_required",
+      "Same-Turn closeout requires the original accepted ordinary completion receipt.", {}, "decision_rejection");
+    if (original.status !== "replayed") return original;
+    const originalValidation = original.validation_receipt as JsonObject | null;
+    const receiptDigest = originalValidation?.validation_declaration_sha256;
+    // Revision-zero receipts may omit the digest. The exact recovered
+    // request already binds the canonical declaration, actor and lease;
+    // later declaration revisions still require the receipt's own digest.
+    const requiresDigest = Number.isSafeInteger(todo.completion_validation_revision) &&
+      Number(todo.completion_validation_revision) > 0;
+    if (original.completion_identity_key !== input.requested_completion_turn_key ||
+        (validationRequired && (originalValidation === null ||
+          typeof originalValidation !== "object" || originalValidation.passed !== true ||
+          (receiptDigest != null && receiptDigest !== validationSha256) ||
+          (requiresDigest && receiptDigest !== validationSha256))) ||
+        (input.validation_receipt !== null &&
+          canonicalAuthoritySha256(input.validation_receipt) !== canonicalAuthoritySha256(original.validation_receipt))) {
+      return terminalFailure("terminal_completion_receipt_mismatch",
+        "Original completion validation and identity must still match the completed Todo.", {}, "decision_rejection");
+    }
+    return original;
+  }
+  return null;
+}
+
 /**
  * Execute complete/supersede as one provider-neutral CAS transaction.
  *
@@ -1239,6 +1301,12 @@ export async function executeCoordinationTodoTerminalLifecycle(
     const receipt = terminalReceipt(input, requestSha);
     const replay = await receipt.read(store);
     if (replay !== null) {
+      // New archived closeout receipts preserve accepted evidence separately
+      // from the historical request fingerprint. Older receipts still replay.
+      if (Object.hasOwn(replay, "evidence")) {
+        const rejection = completionEvidenceRejection(input, replay.evidence);
+        if (rejection !== null) return rejection;
+      }
       // Older accepted lifecycle receipts lack the original quota binding.
       // Add it to this response from verified history; never rewrite the receipt.
       if (replay.status === "replayed" && isLifecycleCloseout(input) &&
@@ -1267,7 +1335,9 @@ export async function executeCoordinationTodoTerminalLifecycle(
     if (!await authoritySourcesCurrent()) return terminalFailure(AUTHORITY_SOURCE_CHANGED.code,
       AUTHORITY_SOURCE_CHANGED.reason, {}, "decision_rejection");
     const observation = await receipt.observe(store);
-    if (observation.kind === "receipt") return observation.result;
+    if (observation.kind === "receipt") return Object.hasOwn(observation.result, "evidence")
+      ? completionEvidenceRejection(input, observation.result.evidence) ?? observation.result
+      : observation.result;
     if (observation.authority.status !== "loaded") {
       return {
         schema_version: COORDINATION_TODO_TERMINAL_LIFECYCLE_RESULT_SCHEMA,
@@ -1360,13 +1430,19 @@ export async function executeCoordinationTodoTerminalLifecycle(
       "Todo requirements changed after Turn admission; reread the task and run a fresh Turn",
       {todo_id: input.todo_id}, "decision_rejection");
   }
+  let terminalUpgradeReceipt: CoordinationTodoTerminalLifecycleResult | null = null;
   if (todo.archive_state !== "active") {
-    return terminalFailure(
-      "todo_archived",
-      "Todo terminal lifecycle requires an active Todo",
-      {},
-      "decision_rejection",
-    );
+    // Ordinary completion may archive this Todo before its original Turn
+    // settles. Only its exact accepted parent permits this second phase;
+    // archive alone grants no new execution or completion authority.
+    terminalUpgradeReceipt = await readSameTurnCompletionReceipt(store, input, normalized, todo);
+    if (terminalUpgradeReceipt === null) {
+      return terminalFailure("todo_archived", "Todo terminal lifecycle requires an active Todo",
+        {}, "decision_rejection");
+    }
+    if (terminalUpgradeReceipt.status !== "replayed") return terminalUpgradeReceipt;
+    const rejection = completionEvidenceRejection(input, todo.evidence);
+    if (rejection !== null) return rejection;
   }
   const authorityTodo = todo;
   let edit: ReturnType<typeof prepareUpdatedTodo> | null = null;
@@ -1607,50 +1683,13 @@ export async function executeCoordinationTodoTerminalLifecycle(
     }
   }
   let completion: ReturnType<typeof reduceTodoCompletionTransaction> | null = null;
-  let terminalUpgradeReceipt: JsonObject | null = null;
   if (input.command === "complete") {
     const validationRequired = todo.completion_validation_required === true;
     const validationSha256 = todo.completion_validation_sha256;
-    if (normalized.operation_identity.kind === "completion_turn" && input.requested_no_followup) {
-      let fence;
-      try {
-        fence = evaluateTodoCompletionFence({schema_version: TODO_COMPLETION_FENCE_REQUEST_SCHEMA,
-          projection_source: "materialized", goal_id: input.goal_id, todo_id: input.todo_id, todo,
-          requested_no_followup: input.requested_no_followup,
-          requested_completion_turn_key: input.requested_completion_turn_key,
-          requested_completion_identity_source: input.requested_completion_identity_source});
-      } catch (error) {
-        return terminalFailure("invalid_todo_completion_transaction",
-          error instanceof Error ? error.message : "Invalid original completion identity");
-      }
-      if (fence.outcome === "continue" && fence.reason === "same_turn_terminal_upgrade") {
-        const originalInput = {...input,
-          ...(normalized.claim_operation_id === undefined ? {} : {claim_operation_id: normalized.claim_operation_id,
-            lease_idempotency_key: normalized.lease_idempotency_key, lease_expected_version: normalized.lease_expected_version}),
-          requested_no_followup: false,
-          operation_id: completionTurnOperationId(input, false)};
-        const original = await terminalReceipt(originalInput, terminalRequestSha(originalInput)).read(store);
-        if (original === null) return terminalFailure("terminal_completion_receipt_required",
-          "Same-Turn closeout requires the original accepted ordinary completion receipt.", {}, "decision_rejection");
-        if (original.status !== "replayed") return original;
-        const originalValidation = original.validation_receipt as JsonObject | null;
-        const receiptDigest = originalValidation?.validation_declaration_sha256;
-        // Revision-zero receipts may omit the digest. The exact recovered
-        // request already binds the canonical declaration, actor and lease;
-        // later declaration revisions still require the receipt's own digest.
-        const requiresDigest = Number.isSafeInteger(todo.completion_validation_revision) &&
-          Number(todo.completion_validation_revision) > 0;
-        if (original.completion_identity_key !== input.requested_completion_turn_key ||
-            (validationRequired && (originalValidation === null ||
-              typeof originalValidation !== "object" || originalValidation.passed !== true ||
-              (receiptDigest != null && receiptDigest !== validationSha256) ||
-              (requiresDigest && receiptDigest !== validationSha256))) ||
-            (input.validation_receipt !== null &&
-              canonicalAuthoritySha256(input.validation_receipt) !== canonicalAuthoritySha256(original.validation_receipt))) {
-          return terminalFailure("terminal_completion_receipt_mismatch",
-            "Original completion validation and identity must still match the completed Todo.", {}, "decision_rejection");
-        }
-        terminalUpgradeReceipt = original;
+    if (terminalUpgradeReceipt === null) {
+      terminalUpgradeReceipt = await readSameTurnCompletionReceipt(store, input, normalized, todo);
+      if (terminalUpgradeReceipt !== null && terminalUpgradeReceipt.status !== "replayed") {
+        return terminalUpgradeReceipt;
       }
     }
     if (validationRequired && !implicitMonitorNoChange &&
@@ -1985,6 +2024,10 @@ export async function executeCoordinationTodoTerminalLifecycle(
     todo_id: input.todo_id,
     command: input.command,
     changed,
+    // Freeze this archived phase's evidence for receipt-only recovery without
+    // adding a fresh authority read or changing legacy operation fingerprints.
+    ...(authorityTodo.archive_state !== "active" && terminalUpgradeReceipt !== null
+      ? {evidence: authorityTodo.evidence ?? null} : {}),
     ...(input.decision_outcome === null ? {} : {decision_outcome: input.decision_outcome}),
     ...(followthrough?.unblock_resume ? {unblock_resume: followthrough.unblock_resume} : {}),
     ...(followthrough?.decision_scope_resolution ? {decision_scope_resolution: followthrough.decision_scope_resolution} : {}),
diff --git a/tests/control_plane/test_quota_settlement_cli.py b/tests/control_plane/test_quota_settlement_cli.py
index a20fdcbab7..3e5518e800 100644
--- a/tests/control_plane/test_quota_settlement_cli.py
+++ b/tests/control_plane/test_quota_settlement_cli.py
@@ -8059,6 +8059,196 @@ def test_completed_todo_reentry_preserves_original_settlement_commands(
     assert settled["interaction_contract"]["agent_channel"]["must_attempt"] is False
 
 
+@pytest.mark.parametrize("goal_commitment", [False, True])
+@pytest.mark.parametrize("provider", ["file", "sqlite"])
+def test_original_turn_closes_automatically_archived_todo_once(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch, provider: str, goal_commitment: bool,
+) -> None:
+    from canonical_authority_fixture import isolate_sqlite_runtime
+
+    # A real validator checks a synthetic result; its counter detects any
+    # repeated validation during closeout or receipt recovery.
+    project, runtime, registry_path = _write_fixture(tmp_path, claimed=False)
+    state_path = _configure_read_only_todo(project)
+    artifact = project / "result.json"
+    artifact.write_text('{"sum":45}\n', encoding="utf-8")
+    counter = project / "validator-count.txt"
+    validator = (
+        "import json; from pathlib import Path; "
+        f"assert json.loads(Path({str(artifact)!r}).read_text()) == {{'sum':45}}; "
+        f"p=Path({str(counter)!r}); "
+        "p.write_text(str(int(p.read_text())+1 if p.exists() else 1))"
+    )
+    argv = quote(json.dumps([sys.executable, "-c", validator]), safe="")
+    state_path.write_text(
+        state_path.read_text(encoding="utf-8").replace(
+            "action_kind=validate ",
+            f"action_kind=validate validation_command_argv={argv} ", 1,
+        ) + "".join(
+            f"\n- [x] [P2] Completed synthetic history {index}.\n"
+            f"  <!-- loopx:todo todo_id=todo_history_{index:04d} status=done "
+            "task_class=advancement_task action_kind=validate -->\n"
+            for index in range(100)
+        ),
+        encoding="utf-8",
+    )
+    if provider == "sqlite":
+        isolate_sqlite_runtime(tmp_path, monkeypatch)
+    _initialize_fixture_authority(project, runtime, provider=provider,
+                                  handoff_mode="hard_lease", leases=[])
+    goal_counter = project / "goal-validator-count.txt"
+    if goal_commitment:
+        # A failed cross-task Goal check cannot replace or block the local
+        # validator, nor become accepted just because this Todo closes.
+        checks = tmp_path / "goal-checks.json"
+        checks.write_text(json.dumps({
+            "scope": {"kind": "goal"}, "objective": "Fulfil the overall commitment",
+            "non_goals": [], "bindings": [], "criteria": [{
+                "id": "overall-commitment", "description": "Overall work remains incomplete",
+                "validation_argv": [sys.executable, "-c",
+                    "from pathlib import Path; "
+                    f"p=Path({str(goal_counter)!r}); "
+                    "p.write_text(str(int(p.read_text())+1 if p.exists() else 1)); "
+                    "raise SystemExit(1)"],
+            }],
+        }), encoding="utf-8")
+        rc, inspected = _run_cli(registry_path, runtime, "goal-acceptance", "inspect",
+                                 "--goal-id", GOAL_ID)
+        assert rc == 0, inspected
+        rc, configured = _run_cli(
+            registry_path, runtime, "goal-acceptance", "configure", "--goal-id", GOAL_ID,
+            "--document", str(checks), "--operation-id", "overall-check",
+            "--expected-provider-revision", inspected["provider_revision"], "--execute",
+        )
+        assert rc == 0, configured
+        rc, failed = _run_cli(registry_path, runtime, "goal-acceptance", "verify",
+                             "--goal-id", GOAL_ID, "--agent-id", AGENT_ID, "--execute")
+        assert rc == 1 and failed["checks_passed"] is False, failed
+        assert goal_counter.read_text() == "1"
+    claim_id = "original-archive-closeout-claim"
+    rc, claim = _run_cli(registry_path, runtime, "todo", "claim",
+                        "--goal-id", GOAL_ID, "--todo-id", TODO_ID,
+                        "--agent-id", AGENT_ID, "--claimed-by", AGENT_ID,
+                        "--claim-operation-id", claim_id)
+    assert rc == 0, claim
+    binding = ("--goal-id", GOAL_ID, "--agent-id", AGENT_ID,
+               "--todo-id", TODO_ID, "--turn-instance-id", TURN_ID)
+    rc, guard = _run_cli(registry_path, runtime, "quota", "should-run",
+                        "--codex-app", *binding)
+    assert rc == 0 and guard["normal_delivery_allowed"], guard
+    identity = guard["heartbeat_receipt"]["settlement_identity"]
+    ordinary_args = ("todo", "complete", *binding, "--claim-operation-id", claim_id,
+                     "--evidence", "artifact:verified-sum")
+    rc, ordinary = _run_cli(registry_path, runtime, *ordinary_args)
+    assert rc == 0 and ordinary["validation_receipt"]["passed"], ordinary
+    rc, archived = _run_cli(registry_path, runtime, "todo", "list",
+                           "--goal-id", GOAL_ID, "--todo-id", TODO_ID,
+                           "--include-archived")
+    assert rc == 0, archived
+    assert archived["todo"]["archive_state"] == "archive"
+    assert archived["todo"]["status"] == "done"
+    completed_at = archived["todo"]["completed_at"]
+    terminal_args = (*ordinary_args, "--no-follow-up")
+    rc, premature = _run_cli(registry_path, runtime, *terminal_args)
+    assert rc == 1, premature
+    assert "requires matching writeback and quota spend receipts" in premature["error"]
+    assert _spend_run_count(runtime) == 0
+
+    rc, written = _run_cli(
+        registry_path, runtime, "refresh-state", *binding,
+        "--classification", "archived_result_verified",
+        "--delivery-batch-scale", "single_surface",
+        "--delivery-outcome", "outcome_progress",
+        "--vision-state", "vision_closed",
+        "--vision-acceptance", "The independently verified sum is durably recorded.",
+        "--no-global-sync", "--suppress-external-sinks",
+    )
+    assert rc == 0, written
+    spend_args = ("quota", "spend-slot", *binding, "--slots", "1",
+                  "--source", "heartbeat", "--execute")
+    for _ in range(2):
+        rc, spent = _run_cli(registry_path, runtime, *spend_args)
+        assert rc == 0, spent
+    validation_runs = counter.read_text()
+    assert validation_runs == "1"
+
+    # A reused validation receipt cannot authorize replacing the artifact that
+    # the archived Todo accepted. Changing only evidence must fail before the
+    # terminal receipt exists and after that receipt can be replayed.
+    replacement_args = tuple(
+        "artifact:unvalidated-replacement" if arg == "artifact:verified-sum" else arg
+        for arg in terminal_args
+    )
+    rc, replacement = _run_cli(registry_path, runtime, *replacement_args)
+    assert rc == 1, replacement
+    assert replacement["error_code"] == "terminal_completion_receipt_mismatch", replacement
+    rc, preserved = _run_cli(registry_path, runtime, "todo", "list",
+                             "--goal-id", GOAL_ID, "--todo-id", TODO_ID,
+                             "--include-archived")
+    assert rc == 0, preserved
+    assert preserved["todo"]["evidence"] == "artifact:verified-sum"
+    assert counter.read_text() == validation_runs
+    assert _spend_run_count(runtime) == 1
+
+    # An unrelated archived task, another actor, and a different claim or Turn
+    # cannot borrow the completed task's original receipt or settlement.
+    attempts = [
+        ("todo", "complete", "--goal-id", GOAL_ID, "--agent-id", AGENT_ID,
+         "--todo-id", "todo_history_0000"),
+        tuple("unregistered-actor" if arg == AGENT_ID else arg for arg in terminal_args),
+        tuple("unrelated-claim" if arg == claim_id else arg for arg in terminal_args),
+        tuple("different-turn" if arg == TURN_ID else arg for arg in terminal_args),
+        (*terminal_args, "--clear-claim"),
+    ]
+    for attempt in attempts:
+        rc, refused = _run_cli(registry_path, runtime, *attempt)
+        assert rc == 1, refused
+        assert _spend_run_count(runtime) == 1
+        assert counter.read_text() == validation_runs
+    assert refused["error_code"] == "coordination_operation_identity_mismatch", refused
+
+    rc, terminal = _run_cli(registry_path, runtime, *terminal_args)
+    assert rc == 0, terminal
+    assert terminal["completion_continuation"] == "no_followup"
+    assert terminal["completion_recovery"] == "same_turn_terminal_closeout"
+    assert terminal["completion_identity_key"] == ordinary["completion_identity_key"]
+    assert terminal["validation_receipt"] == ordinary["validation_receipt"]
+    assert [row["step_kind"] for row in terminal["settlement_result"]["receipts"]] == [
+        "validation", "durable_writeback", "quota_spend", "terminal_closeout",
+    ]
+    evidence_index = terminal_args.index("--evidence")
+    omitted_evidence = terminal_args[:evidence_index] + terminal_args[evidence_index + 2:]
+    empty_evidence = (*terminal_args[:evidence_index + 1], "", *terminal_args[evidence_index + 2:])
+    for args in (omitted_evidence, empty_evidence):
+        rc, refused = _run_cli(registry_path, runtime, *args)
+        assert rc == 1 and "requires --note or --evidence" in refused["error"], refused
+    completion_note = ("--note", "Close the original independently verified result.")
+    for args in (terminal_args, ordinary_args,
+                 (*omitted_evidence, *completion_note), (*empty_evidence, *completion_note)):
+        rc, replay = _run_cli(registry_path, runtime, *args)
+        assert rc == 0 and replay["idempotent_replay"], replay
+    rc, replacement = _run_cli(registry_path, runtime, *replacement_args)
+    assert rc == 1, replacement
+    assert replacement["error_code"] == "terminal_completion_receipt_mismatch", replacement
+    rc, final = _run_cli(registry_path, runtime, "todo", "list",
+                        "--goal-id", GOAL_ID, "--todo-id", TODO_ID,
+                        "--include-archived")
+    assert rc == 0, final
+    assert final["todo"]["archive_state"] == "archive"
+    assert final["todo"]["completed_at"] == completed_at
+    assert final["todo"]["evidence"] == "artifact:verified-sum"
+    assert final["todo"]["completion_turn_key"] == identity["effect_id"]
+    assert counter.read_text() == validation_runs
+    assert _spend_run_count(runtime) == 1
+    assert _classification_count(runtime, "archived_result_verified") == 1
+    if goal_commitment:
+        rc, overall = _run_cli(registry_path, runtime, "goal-acceptance", "inspect",
+                               "--goal-id", GOAL_ID)
+        assert rc == 0, overall
+        assert overall["goal_acceptance_contract"]["status"] == "failed"
+        assert goal_counter.read_text() == "1"
+
+
 @pytest.mark.parametrize("provider", ["file", "sqlite"])
 def test_read_only_settlement_omits_non_causal_delivery_workspace(
     tmp_path: Path,