diff --git a/packages/plugins/plugin-approvals/src/stranded-resubmit-second-door.test.ts b/packages/plugins/plugin-approvals/src/stranded-resubmit-second-door.test.ts
new file mode 100644
index 0000000000..fb3c789581
--- /dev/null
+++ b/packages/plugins/plugin-approvals/src/stranded-resubmit-second-door.test.ts
@@ -0,0 +1,365 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * #17601 PROBE — RECORDED DEFECT, not endorsed behaviour.
+ *
+ * ⛔ Nothing here asserts that a second `resubmit` on a stranded-resubmit strand
+ * is CORRECT. This file records what the door does today, because the card it
+ * came from was a hypothesis derived from reading and explicitly not executed,
+ * and the next reader of `continueRestoredRun` should not have to re-derive the
+ * same chain from the same lines. Whether the door should be closed, or the
+ * discriminator's prose scoped instead, is a ruling-grade question the card
+ * routed to the decision box. ⇒ When that ruling lands and the repair changes
+ * this behaviour, THIS PIN is the thing to update, and its update is the
+ * repair's evidence.
+ *
+ * ## The hypothesis, and what it tested
+ *
+ * `resolveRecordedContinuation` discriminates the two continuation issuers of a
+ * `returned` row by the presence of an `action: 'resubmit'` audit row, and
+ * argues the discriminator is *"exact and structural"* on three clauses:
+ * `action: 'resubmit'` has exactly one writer in the file, it is inserted
+ * before that resume, and a resubmit opens the next round as a NEW row — *"so
+ * at most one such action row exists per request"*.
+ *
+ * ⭐ MEASURED: the third clause does not hold on the stranded-resubmit strand.
+ * A `resubmit` whose resume strands writes its audit row and opens NO new round
+ * (the row stays `returned`), so once an operator re-arms the pause with
+ * `restoreConsumedSuspension` — the state `continueRestoredRun` exists to serve
+ * — a second `resubmit` by the same submitter passes all five door guards and
+ * writes a SECOND `action: 'resubmit'` row on the same request.
+ *
+ * ## The bound on it, measured too, so the finding is not read wider than it is
+ *
+ * The discriminator's own READ is a presence check (`limit: 1`), so two rows
+ * decide exactly as one does: MEASUREMENT C drives the resolver on the doubled
+ * row and it still answers `resubmit`. ⇒ What is falsified is the stated
+ * invariant and the audit trail's one-row-per-advancement shape, ⛔ not (on
+ * today's code) the edge the repair verb reads.
+ */
+
+import { describe, it, expect } from 'vitest';
+import { AutomationEngine, InMemorySuspendedRunStore } from '@objectstack/service-automation';
+// [#4550] The engine double routes its write verbs through ObjectQL's OWN
+// dispatch predicates rather than a hand-mirrored copy.
+import { assertEngineDeleteDispatch, assertEngineUpdateDispatch } from '@objectstack/objectql';
+import { APPROVAL_REVISE_NODE_TYPE } from '@objectstack/spec/automation';
+import { strandedDecisionDetails } from '@objectstack/types';
+import { ApprovalService } from './approval-service.js';
+import { registerApprovalNode } from './approval-node.js';
+
+const SYSTEM_CTX = { isSystem: true, positions: [], permissions: [] } as any;
+const asUser = (userId: string) =>
+  ({ isSystem: false, userId, positions: [], permissions: [] }) as any;
+const noopLogger = { info() {}, warn() {}, error() {}, debug() {} };
+
+/** In-memory ObjectQL stand-in for the approvals tables. */
+function makeFakeEngine() {
+  const tables = new Map<string, any[]>();
+  const rows = (o: string) => (tables.get(o) ?? (tables.set(o, []), tables.get(o)!));
+  // The lever that makes a REAL strand reachable from a test: fail the very next
+  // insert into one table, once. The approval node's executor opens the next
+  // round by inserting a `sys_approval_request`, so failing that insert strands
+  // the resume the same way a downstream node's throw does — `RESUME_FAILED`
+  // with `repairable: true`, the suspension already consumed.
+  let failNextInsertOn: string | undefined;
+  const matches = (row: any, where: any) => Object.entries(where ?? {}).every(([k, v]) => {
+    if (k.startsWith('$')) throw new Error(`fake engine: unsupported filter operator ${k}`);
+    if (v && typeof v === 'object' && '$in' in (v as any)) return (v as any).$in.includes(row[k]);
+    if (v && typeof v === 'object' && '$ne' in (v as any)) return row[k] !== (v as any).$ne;
+    return row[k] === v;
+  });
+  return {
+    tables,
+    set failNextInsert(object: string | undefined) { failNextInsertOn = object; },
+    get failNextInsert() { return failNextInsertOn; },
+    async find(object: string, opts: any = {}) {
+      const where = opts.where ?? opts.filter ?? {};
+      const out = rows(object).filter(r => matches(r, where));
+      // ⚠️ `orderBy` is honoured, and that is load-bearing rather than polish:
+      // `assertLatestForRun` selects the newest request with
+      // `orderBy [{field:'created_at', order:'desc'}], limit 1`. A double that
+      // ignored it returned the OLDEST row, so the guard passed on every input
+      // and a pin naming it would have measured nothing — the phantom-check
+      // shape. SortNode's key is `order`, not `direction`.
+      if (Array.isArray(opts.orderBy)) {
+        for (const sort of [...opts.orderBy].reverse()) {
+          const field = sort?.field;
+          if (!field) continue;
+          const dir = sort?.order === 'desc' ? -1 : 1;
+          out.sort((a, b) => (a[field] < b[field] ? -1 : a[field] > b[field] ? 1 : 0) * dir);
+        }
+      }
+      // The caller's bound is honoured by PRESENCE, never truthiness.
+      const start = opts.offset ?? 0;
+      const page = typeof opts.limit === 'number' ? out.slice(start, start + opts.limit) : out.slice(start);
+      return page.map(r => ({ ...r }));
+    },
+    async insert(object: string, data: any) {
+      if (failNextInsertOn === object) {
+        failNextInsertOn = undefined;
+        throw new Error(`injected one-shot insert failure on ${object}`);
+      }
+      rows(object).push({ ...data }); return { ...data };
+    },
+    async update(object: string, data: any, options?: any) {
+      const dispatch = assertEngineUpdateDispatch(data, options);
+      const table = rows(object);
+      if (dispatch.kind === 'multi') {
+        let n = 0;
+        for (let i = 0; i < table.length; i++) {
+          if (matches(table[i], options?.where)) { table[i] = { ...table[i], ...data }; n++; }
+        }
+        return { updated: n };
+      }
+      const i = table.findIndex(r => r.id === dispatch.id);
+      if (i >= 0) table[i] = { ...table[i], ...data };
+      return i >= 0 ? { ...table[i] } : null;
+    },
+    async delete(object: string, options?: any) {
+      const dispatch = assertEngineDeleteDispatch(options);
+      const table = rows(object);
+      if (dispatch.kind === 'multi') {
+        const survivors = table.filter(r => !matches(r, options?.where));
+        const deleted = table.length - survivors.length;
+        table.splice(0, table.length, ...survivors);
+        return { deleted };
+      }
+      const i = table.findIndex(r => r.id === dispatch.id);
+      if (i >= 0) table.splice(i, 1);
+      return { id: dispatch.id };
+    },
+  };
+}
+
+/** ADR-0044 revise window: send-back parks the run at the `approval_revise` node. */
+const REVISE_FLOW = {
+  name: 'revise_flow', label: 'Revise Flow', type: 'autolaunched',
+  nodes: [
+    { id: 'start', type: 'start', label: 'Start' },
+    { id: 'review', type: 'approval', label: 'Review', config: { approvers: [{ type: 'user', value: 'u1' }] } },
+    { id: 'wait_revision', type: APPROVAL_REVISE_NODE_TYPE, label: 'Awaiting Revision' },
+    { id: 'on_approved', type: 'mark', label: 'Approved' },
+    { id: 'on_rejected', type: 'mark', label: 'Rejected' },
+    { id: 'end', type: 'end', label: 'End' },
+  ],
+  edges: [
+    { id: 'e1', source: 'start', target: 'review' },
+    { id: 'e2', source: 'review', target: 'on_approved', label: 'approve' },
+    { id: 'e3', source: 'review', target: 'on_rejected', label: 'reject' },
+    { id: 'e4', source: 'review', target: 'wait_revision', label: 'revise' },
+    { id: 'e5', source: 'wait_revision', target: 'review', label: 'resubmit', type: 'back' },
+    { id: 'e6', source: 'on_approved', target: 'end' },
+    { id: 'e7', source: 'on_rejected', target: 'end' },
+  ],
+};
+
+/**
+ * One live process per scenario — real engine, real approval node, real
+ * approvals service. Each scenario owns its own tables, so no leg can select a
+ * row another leg left behind.
+ */
+function scenario() {
+  const marks: string[] = [];
+  const data = makeFakeEngine();
+  const service = new ApprovalService({ engine: data as any, logger: noopLogger });
+  const automation = new AutomationEngine(noopLogger as any, new InMemorySuspendedRunStore());
+  registerApprovalNode(automation, service, noopLogger as any);
+  automation.registerNodeExecutor({
+    type: 'mark',
+    async execute(node: any) { marks.push(node.id); return { success: true }; },
+  } as never);
+  automation.registerFlow('revise_flow', REVISE_FLOW as never);
+  service.attachAutomation(automation);
+
+  const countActions = async (requestId: string, action: string) =>
+    (await data.find('sys_approval_action', { where: { request_id: requestId, action } })).length;
+  const runRows = async (runId: string) =>
+    await data.find('sys_approval_request', { where: { flow_run_id: runId } });
+  const rowOf = async (id: string) =>
+    (await data.find('sys_approval_request', { where: { id } }))[0];
+  const parkedAt = async (runId: string) =>
+    (await automation.listSuspendedRunsDurable()).find((r: any) => String(r.runId) === String(runId))?.nodeId;
+  const tryResubmit = (requestId: string, userId: string) => service
+    .resubmit(requestId, { actorId: userId } as any, asUser(userId))
+    .then((value: any) => ({ ok: true as const, value }), (e: Error) => ({ ok: false as const, message: e.message }));
+
+  /**
+   * ⛔ PRECONDITION, asserted step by step. A probe that never reached the
+   * stranded state is a probe that measured NOTHING about the second call, so
+   * every step of getting there is a hard assertion rather than a setup line.
+   */
+  async function strandAResubmit(recordId: string) {
+    await automation.execute('revise_flow', {
+      object: 'crm_deal', record: { id: recordId, amount: 100 }, userId: 'submitter',
+    } as never);
+    const req: any = (await data.find('sys_approval_request', {
+      where: { record_id: recordId, status: 'pending' },
+    }))[0];
+    expect(req?.flow_node_id, 'PRECONDITION: parked at the approval node').toBe('review');
+    const runId = String(req.flow_run_id);
+
+    await service.sendBack(req.id, { actorId: 'u1', comment: 'redo' } as any, SYSTEM_CTX);
+    expect(await parkedAt(runId), 'PRECONDITION: the send-back parked the run at the revise window')
+      .toBe('wait_revision');
+    expect((await rowOf(req.id)).status, 'PRECONDITION: and the row reads `returned`').toBe('returned');
+
+    // Strand the resubmit: the back-edge re-enters `review`, whose executor
+    // opens round 2 by inserting a request — fail that insert, once.
+    data.failNextInsert = 'sys_approval_request';
+    const stranded = await service
+      .resubmit(req.id, { actorId: 'submitter' } as any, asUser('submitter'))
+      .then(() => null, (e: Error) => e);
+    expect(stranded?.message, 'PRECONDITION: a REAL stranded resubmit').toMatch(/^RESUME_FAILED/);
+    expect(strandedDecisionDetails(stranded)?.repairable,
+      'PRECONDITION: the repairable strand this card is about').toBe(true);
+    expect(data.failNextInsert, 'PRECONDITION: the injected failure fired and was consumed').toBeUndefined();
+
+    // ⭐ The card's own description of the third strand shape, measured rather
+    // than assumed: `returned`, no newer row, one audit row, pause consumed.
+    expect((await rowOf(req.id)).status, 'PRECONDITION: the row is left `returned`').toBe('returned');
+    expect(await runRows(runId), 'PRECONDITION: with NO newer row on the run').toHaveLength(1);
+    expect(await countActions(req.id, 'resubmit'),
+      'PRECONDITION: exactly one `action: resubmit` row so far').toBe(1);
+    expect(await automation.hasSuspendedRun(runId),
+      'PRECONDITION: the stranded resume consumed the suspension').toBe(false);
+
+    return { req, runId };
+  }
+
+  return { data, service, automation, marks, countActions, runRows, rowOf, parkedAt, tryResubmit, strandAResubmit };
+}
+
+describe('#17601 — a second `resubmit` on a stranded-resubmit strand', () => {
+  it('MEASUREMENT A — re-armed: every door guard passes and a SECOND `resubmit` audit row lands', async () => {
+    const s = scenario();
+    const { req, runId } = await s.strandAResubmit('d1');
+
+    // The re-arm the card's `assertRunResumable → passes via hasSuspendedRun`
+    // row depends on. This is not an exotic state: it is exactly the state
+    // `restoreConsumedSuspension` puts a stranded run into, and the one
+    // `continueRestoredRun` was built to serve.
+    const rearmed = await s.automation.restoreConsumedSuspension(runId, { requestedBy: 'ops' });
+    expect(rearmed.restored, 'PRECONDITION: the pause really is back').toBe(true);
+    expect(await s.automation.hasSuspendedRun(runId)).toBe(true);
+    expect(await s.parkedAt(runId), 'PRECONDITION: re-armed at the revise window').toBe('wait_revision');
+
+    // ── ⛔ FIRING CONTROLS. Each guard the card reads as "passes" is shown able
+    // to REFUSE on this very row, in this very state — otherwise "it passed" is
+    // indistinguishable from "it was never consulted".
+    const foreign = await s.tryResubmit(req.id, 'u1');
+    expect(foreign.ok, 'CONTROL: the submitter-only guard still fires').toBe(false);
+    expect(foreign.ok === false && foreign.message).toMatch(/^FORBIDDEN: only the submitter may resubmit/);
+
+    const requests = s.data.tables.get('sys_approval_request')!;
+    const before = requests.length;
+    // A colliding PENDING request on the same record, on another run so the
+    // supersede guard cannot be the thing that speaks.
+    requests.push({
+      id: 'areq_collider', object_name: 'crm_deal', record_id: 'd1', status: 'pending',
+      flow_run_id: 'run_other', created_at: '2020-01-01T00:00:00.000Z',
+    });
+    const collided = await s.tryResubmit(req.id, 'submitter');
+    expect(collided.ok, 'CONTROL: the collision check still fires').toBe(false);
+    expect(collided.ok === false && collided.message).toMatch(/^DUPLICATE_REQUEST/);
+    requests.pop();
+
+    // A newer row on the SAME run, so `assertLatestForRun` is the speaker.
+    requests.push({
+      id: 'areq_newer', object_name: 'crm_deal', record_id: 'd1', status: 'pending',
+      flow_run_id: runId, created_at: '2099-01-01T00:00:00.000Z',
+    });
+    const superseded = await s.tryResubmit(req.id, 'submitter');
+    expect(superseded.ok, 'CONTROL: the supersede guard still fires').toBe(false);
+    expect(superseded.ok === false && superseded.message)
+      .toBe('INVALID_STATE: a newer approval request supersedes this one');
+    requests.pop();
+    expect(requests.length, 'CONTROL: both control rows removed — the state is the one under test')
+      .toBe(before);
+    expect(await s.countActions(req.id, 'resubmit'),
+      'CONTROL: and not one refusal wrote an audit row').toBe(1);
+
+    // ── ⭐ THE MEASUREMENT. Same submitter, nothing else changed.
+    const second = await s.tryResubmit(req.id, 'submitter');
+
+    expect(second.ok, '⭐ the second `resubmit` is ADMITTED — the hypothesis reproduces').toBe(true);
+    expect(second.ok === true && second.value.resumed, 'and the run moved').toBe(true);
+    expect(second.ok === true && second.value.runId).toBe(runId);
+
+    // ⭐ The falsified clause: `action: 'resubmit'` is written twice for one
+    // request, where `resolveRecordedContinuation` records that at most one
+    // such row exists per request.
+    expect(await s.countActions(req.id, 'resubmit'),
+      '⭐ TWO `action: resubmit` rows on one request').toBe(2);
+
+    // What did and did not follow from it. Round 2 opened exactly ONCE — the
+    // flow is not doubly advanced — and the row the second call acted on is
+    // still the `returned` round-1 row, untouched.
+    expect(await s.runRows(runId), 'round 2 opened, once').toHaveLength(2);
+    expect((await s.rowOf(req.id)).status, 'the round-1 row is still `returned`').toBe('returned');
+    expect(await s.parkedAt(runId), 'and the run is parked back at the approval node').toBe('review');
+    expect(s.marks, 'no downstream mark node ran — the back-edge re-parks at `review`').toEqual([]);
+  });
+
+  it('MEASUREMENT B — CONTROL, un-re-armed: `assertRunResumable` refuses before anything is written', async () => {
+    // The same second call with the ONE difference that matters: no re-arm, so
+    // the suspension the stranded resume consumed is still gone. This is what
+    // keeps MEASUREMENT A from being a statement about `resubmit` in general —
+    // the re-armed pause is the specific thing that opens the door.
+    const s = scenario();
+    const { req, runId } = await s.strandAResubmit('d2');
+
+    const second = await s.tryResubmit(req.id, 'submitter');
+    expect(second.ok, 'refused').toBe(false);
+    expect(second.ok === false && second.message).toMatch(/^RESUME_TARGET_LOST: the flow run/);
+
+    // ⭐ And the refusal lands BEFORE the insert, which is why the audit row
+    // count is the discriminating reading between the two measurements.
+    expect(await s.countActions(req.id, 'resubmit'), 'still exactly one audit row').toBe(1);
+    expect(await s.runRows(runId), 'and no round 2').toHaveLength(1);
+  });
+
+  it('MEASUREMENT C — the BOUND: the doubled row still rebuilds as `resubmit`, not as something else', async () => {
+    // ⛔ Keeps the finding from being read wider than it is. The discriminator's
+    // read is a PRESENCE check (`limit: 1`), so two rows answer exactly as one
+    // does. What the doubling falsifies is the stated invariant and the audit
+    // trail's one-row-per-advancement shape — ⛔ not, on today's code, the edge
+    // the repair verb reads.
+    const s = scenario();
+    const { req, runId } = await s.strandAResubmit('d3');
+    await s.automation.restoreConsumedSuspension(runId, { requestedBy: 'ops' });
+    const second = await s.tryResubmit(req.id, 'submitter');
+    expect(second.ok, 'PRECONDITION: the doubling happened').toBe(true);
+    expect(await s.countActions(req.id, 'resubmit')).toBe(2);
+
+    // Force the rebuild path: strip the journal the failing door wrote, which
+    // is also what a run stranded before the journal shipped looks like.
+    const config = JSON.parse((await s.rowOf(req.id)).node_config_json);
+    expect(config.__strandedContinuation, 'PRECONDITION: the door journalled it').toBeTruthy();
+    delete config.__strandedContinuation;
+    await s.data.update('sys_approval_request', {
+      id: req.id, node_config_json: JSON.stringify(config),
+    }, { context: SYSTEM_CTX });
+
+    const rebuilt = await (s.service as any).resolveRecordedContinuation(await s.rowOf(req.id), req.id);
+    expect(rebuilt.source).toBe('reconstructed');
+    expect(rebuilt.signal.decision, 'two rows decide as one row does').toBe('resubmit');
+    expect(rebuilt.signal.branchLabel).toBe('resubmit');
+
+    // REVERSE CONTROL: the same resolver on a `returned` row with NO resubmit
+    // action row answers `revise` — so the line above is a reading, not a
+    // constant this resolver returns for every input.
+    const bare = scenario();
+    await bare.automation.execute('revise_flow', {
+      object: 'crm_deal', record: { id: 'd4', amount: 4 }, userId: 'submitter',
+    } as never);
+    const other: any = (await bare.data.find('sys_approval_request', {
+      where: { record_id: 'd4', status: 'pending' },
+    }))[0];
+    await bare.service.sendBack(other.id, { actorId: 'u1', comment: 'redo' } as any, SYSTEM_CTX);
+    expect(await bare.countActions(other.id, 'resubmit'), 'CONTROL: no resubmit row on this one').toBe(0);
+    const rebuiltBare = await (bare.service as any)
+      .resolveRecordedContinuation(await bare.rowOf(other.id), other.id);
+    expect(rebuiltBare.signal.decision, 'CONTROL: no resubmit row ⇒ the send-back').toBe('revise');
+  });
+});
diff --git a/scripts/engine-double-contract.pinned.json b/scripts/engine-double-contract.pinned.json
index 30076fb4ca..b70be27081 100644
--- a/scripts/engine-double-contract.pinned.json
+++ b/scripts/engine-double-contract.pinned.json
@@ -2296,6 +2296,16 @@
       "verb": "update",
       "pinned": 1
     },
+    {
+      "file": "packages/plugins/plugin-approvals/src/stranded-resubmit-second-door.test.ts",
+      "verb": "delete",
+      "pinned": 1
+    },
+    {
+      "file": "packages/plugins/plugin-approvals/src/stranded-resubmit-second-door.test.ts",
+      "verb": "update",
+      "pinned": 1
+    },
     {
       "file": "packages/plugins/plugin-approvals/src/stranded-run-repairability.test.ts",
       "verb": "delete",