diff --git a/.changeset/19452-batch-causal-row-located-by-fault.md b/.changeset/19452-batch-causal-row-located-by-fault.md
new file mode 100644
index 00000000000..d11523bb1a2
--- /dev/null
+++ b/.changeset/19452-batch-causal-row-located-by-fault.md
@@ -0,0 +1,25 @@
+---
+'@objectstack/metadata-protocol': patch
+---
+
+fix(metadata-protocol): a stopped or rolled-back bulk batch names the row that actually failed
+
+`reconcileStoppedBatch` and `buildRolledBackBatchResponse` — the two builders every one of the three bulk-write faces (`batchData`, `updateManyData`, `deleteManyData`) reports through — located the causal row with `findIndex(r => !r.success)`. That encoded an invariant: **`!success` means this row failed, and it carries `errors[0]`.**
+
+That invariant stopped holding when a matched-but-deliberately-not-removed row started answering `success: false` with no `errors` entry — correctly, because a surviving record is an outcome, not a fault. The locator could then land on that survivor, `errors?.[0]?.message` was `undefined`, and the message named the **wrong index** while calling the real error — sitting in the same array — 「unknown error」.
+
+Measured on the unfixed tree:
+
+- non-atomic `deleteMany ['survivor', 'missing', 'other']` — the un-attempted row answered `NOT_ATTEMPTED` *"record 0 failed — unknown error; the batch stopped there. …"* while record **1** is what threw;
+- atomic `[t1, survivor, t3]` — the rolled-back rows answered `ROLLED_BACK` *"record 1 failed — unknown error"* for a row that **survived**;
+- atomic `[t3, survivor, missing, t2]` — `ROLLED_BACK` said *"record 1 failed — unknown error"* and `NOT_ATTEMPTED` said *"atomic batch aborted by record 1"*, both naming the survivor while record **2** threw.
+
+Both builders now share one locator, `locateBatchCause`, which finds the row by its recorded **fault** — the row's `errors[]` entry. That is the one per-row value whose declared meaning is a failure: `BatchOperationResultSchema.errors` is documented as *"Array of errors if operation failed"*, and the v17 migration entry publishes `row.errors?.[0]?.message` / `.code` to consumers as exactly that read. Its codes are drawn from the closed `StandardErrorCode ∪ ERROR_CODE_LEDGER` vocabulary, so an unregistered code fails `BatchOperationResultSchema.parse` — giving a non-fault ending an `errors[]` entry is a ledger widening in `packages/spec`, not something a call site can do on its own. `ApiError.message` is required, so a located cause always has text and the 「unknown error」 fallback is **deleted** rather than merely unreached.
+
+The scan runs from the end of the attempted rows, because a run ends *at* the row it stops on: every stop is a `break` in a loop's `catch`, immediately after that row was pushed. A fault that does not stop the run (the `Unknown operation:` arm records one and keeps going) therefore cannot shadow the row that did.
+
+One ending has no fault to quote at all — an atomic batch aborted by a lone survivor, where `runAtomicBatch` rolls back on `failed > 0` and nothing ever threw. The rolled-back rows now read *"record 1 did not succeed"*: the row that stopped the batch committing, named as what it is rather than as a failure with an unknown cause.
+
+No envelope field, per-row code, status or count changes; `succeeded` and `failed` still partition `results`. What changes is which row two message strings name, and both of them stop inventing an error that is not there. Clients branch on `errors[0].code`, which is unchanged — the row classification itself was never wrong.
+
+`Clause-②: no` — nothing authorable moves: no `packages/spec` key, export, accept set or stored shape changes, and the per-row code vocabulary is untouched.
diff --git a/packages/metadata-protocol/src/protocol.batch-causal-row.test.ts b/packages/metadata-protocol/src/protocol.batch-causal-row.test.ts
new file mode 100644
index 00000000000..1aef2450f56
--- /dev/null
+++ b/packages/metadata-protocol/src/protocol.batch-causal-row.test.ts
@@ -0,0 +1,324 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * [#19452] A stopped or rolled-back bulk batch must name the row that actually
+ * failed — and must never call a real error 「unknown」 while it is sitting in
+ * the same array.
+ *
+ * `reconcileStoppedBatch` and `buildRolledBackBatchResponse` located the causal
+ * row with `findIndex(r => !r.success)`, which encoded the invariant
+ * **`!success` ⇒ this row failed and carries `errors[0]`**. #19412 broke that
+ * invariant deliberately and correctly: a row that MATCHED and was NOT removed
+ * (`IDataEngine.delete` answering the count arm's `0`) now reports
+ * `success: false` with ⛔ no `errors` entry, because a surviving record is an
+ * OUTCOME, not a fault.
+ *
+ * ⇒ the locator landed on that survivor, `errors?.[0]?.message` was
+ * `undefined`, and the message named the WRONG index while falling back to
+ * 「unknown error」. Measured on the unfixed tree, with the harness below:
+ *
+ *   deleteMany ['t1'(survives), 'missing'(throws), 't3']
+ *     -> results[2] NOT_ATTEMPTED "record 0 failed — unknown error; the batch
+ *        stopped there. ..."            ⚠️ record 1 is what threw
+ *   batchData atomic [t1, t2(survives), t3]
+ *     -> results[0] ROLLED_BACK "record 1 failed — unknown error"
+ *                                       ⚠️ record 1 SURVIVED; nothing failed
+ *
+ * The negative case — a batch whose FIRST non-success row is a survivor — is
+ * what these pins exist for. The ordinary batches at the bottom are the
+ * positive controls: the same assertions on a run with no survivor in it, so a
+ * locator that simply stopped naming anything cannot pass this file.
+ */
+
+import { describe, it, expect, vi } from 'vitest';
+import { assertEngineDeleteDispatch, assertEngineUpdateDispatch, assertEngineFindOnePredicate } from '@objectstack/metadata-core';
+import { ObjectStackProtocolImplementation } from './protocol.js';
+
+const SCHEMA = {
+    name: 'showcase_private_note',
+    fields: {
+        title: { name: 'title', type: 'text' },
+    },
+};
+
+/** The row the fake engine rejects — a classified failure, as `toRowApiError` expects. */
+const POISON = '__invalid__';
+
+function validationFailure(): Error {
+    const err: any = new Error('title is invalid');
+    err.code = 'VALIDATION_FAILED';
+    err.status = 400;
+    return err;
+}
+
+/**
+ * In-memory store with real snapshot/rollback transaction semantics — the same
+ * harness shape the #4620 / #4793 / #7539 suites use, so every row asserted
+ * here is produced by the actual loops, builders and rollback classifier.
+ *
+ * `delete` speaks the COUNT arm of `IDataEngine.delete`: an unknown id keeps
+ * the contract's `false` (which the loop turns into a thrown
+ * `RECORD_NOT_FOUND`), the nominated `survivor` matches and is deliberately
+ * kept (`0`), everything else really goes (`1`).
+ */
+function makeCountingEngine(survivor: string) {
+    const rows = new Map<string, any>([
+        ['t1', { id: 't1', title: 'stored one' }],
+        ['t2', { id: 't2', title: 'stored two' }],
+        ['t3', { id: 't3', title: 'stored three' }],
+    ]);
+    const handle = { id: 'trx-1' };
+
+    const insert = vi.fn(async (_object: string, data: any) => {
+        if (data?.title === POISON) throw validationFailure();
+        const rec = { id: data.id ?? `new-${rows.size + 1}`, ...data };
+        rows.set(rec.id, rec);
+        return rec;
+    });
+    const update = vi.fn(async (_object: string, data: any, options?: any) => {
+        assertEngineUpdateDispatch(data, options);
+        const id = options?.where?.id;
+        if (data?.title === POISON) throw validationFailure();
+        const next = { ...rows.get(id), ...data };
+        rows.set(id, next);
+        return next;
+    });
+    const del = vi.fn(async (_object: string, options?: any) => {
+        assertEngineDeleteDispatch(options);
+        const id = options?.where?.id;
+        if (!rows.has(id)) return false;   // [#4435] the positive not-found value
+        if (id === survivor) return 0;     // [#19412] matched, deliberately NOT removed
+        rows.delete(id);
+        return 1;
+    });
+    const findOne = vi.fn(async (_object: string, options?: any) => {
+        assertEngineFindOnePredicate(_object, options);
+        return rows.get(options?.where?.id) ?? null;
+    });
+
+    const engine: any = {
+        registry: { getObject: (n: string) => (n === 'showcase_private_note' ? SCHEMA : undefined) },
+        insert,
+        update,
+        delete: del,
+        findOne,
+        getDefaultDriverName: () => 'default',
+        getDriverByName: () => ({ beginTransaction: async () => handle }),
+        transaction: vi.fn(async (callback: (ctx: any) => Promise<any>, baseContext?: any) => {
+            const snapshot = new Map(rows);
+            try {
+                return await callback({ ...(baseContext ?? {}), transaction: handle });
+            } catch (err) {
+                rows.clear();
+                for (const [k, v] of snapshot) rows.set(k, v);
+                throw err;
+            }
+        }),
+    };
+    return { engine, rows, insert, update, del, findOne };
+}
+
+const codesOf = (res: any): Array<string | undefined> =>
+    res.results.map((r: any) => r.errors?.[0]?.code);
+
+/**
+ * The card's acceptance, as ONE assertion usable on every message that names a
+ * causal row: it names the row that really failed, it carries that row's own
+ * error text, and it never says 「unknown error」 while that text exists.
+ *
+ * `causalIndex` is read from the response rather than hard-coded, so the pin
+ * asserts an AGREEMENT between the message and the rows beside it.
+ */
+function expectAttributedTo(message: string, res: any, causalIndex: number) {
+    const causalText = res.results[causalIndex].errors[0].message;
+    expect(causalText).toBeTruthy();
+    expect(message).toContain(`record ${causalIndex} failed`);
+    expect(message).toContain(causalText);
+    expect(message).not.toContain('unknown error');
+}
+
+describe('[#19452] a stopped batch attributes the stop to the row that threw, not to a survivor', () => {
+    it('deleteManyData: a leading survivor does not become the cause', async () => {
+        const t = makeCountingEngine('t1');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.deleteManyData({
+            object: 'showcase_private_note',
+            ids: ['t1', 'definitely_missing', 't3'],
+        } as any);
+
+        // Row 0 survived (no `errors`), row 1 threw, row 2 was never attempted.
+        expect(codesOf(res)).toEqual([undefined, 'RECORD_NOT_FOUND', 'NOT_ATTEMPTED']);
+        expect(res.results[0]).toMatchObject({ id: 't1', success: false });
+        expect(res.results[0].errors).toBeUndefined();   // why `!success` lies here
+        expect(t.rows.has('t1')).toBe(true);             // it really is still there
+
+        // Pre-fix: "record 0 failed — unknown error; the batch stopped there. ..."
+        const message = res.results[2].errors[0].message;
+        expectAttributedTo(message, res, 1);
+        expect(message).not.toContain('record 0');
+        expect(message).toContain('continueOnError');
+    });
+
+    it('batchData delete: the same run through the other non-atomic face', async () => {
+        const t = makeCountingEngine('t1');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.batchData({
+            object: 'showcase_private_note',
+            request: { operation: 'delete', records: [{ id: 't1' }, { id: 'definitely_missing' }, { id: 't3' }] },
+        } as any);
+
+        expect(codesOf(res)).toEqual([undefined, 'RECORD_NOT_FOUND', 'NOT_ATTEMPTED']);
+        expect(res.results[0].errors).toBeUndefined();
+
+        const message = res.results[2].errors[0].message;
+        expectAttributedTo(message, res, 1);
+        expect(message).not.toContain('record 0');
+    });
+});
+
+describe('[#19452] a rolled-back atomic batch attributes the abort to the row that threw', () => {
+    /**
+     * One run that reaches all THREE message sites: a committed-then-undone
+     * row, a survivor, the row that threw, and a row never reached. The two
+     * interpolations of the causal index get SEPARATE tests, so neither can
+     * mask the other's reading when this file runs red.
+     */
+    const atomicRunReachingBothInterpolations = async () => {
+        const t = makeCountingEngine('t1');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.batchData({
+            object: 'showcase_private_note',
+            request: {
+                operation: 'delete',
+                records: [{ id: 't3' }, { id: 't1' }, { id: 'definitely_missing' }, { id: 't2' }],
+                options: { atomic: true },
+            },
+        } as any);
+
+        expect(res).toMatchObject({ success: false, total: 4, succeeded: 0, failed: 4 });
+        expect(codesOf(res)).toEqual(['ROLLED_BACK', undefined, 'RECORD_NOT_FOUND', 'NOT_ATTEMPTED']);
+        return { t, res };
+    };
+
+    it('the ROLLED_BACK message names record 2, not the survivor at record 1', async () => {
+        const { t, res } = await atomicRunReachingBothInterpolations();
+
+        // Pre-fix: "record 1 failed — unknown error" — record 1 SURVIVED.
+        expectAttributedTo(res.results[0].errors[0].message, res, 2);
+        expect(res.results[0].errors[0].message).not.toContain('record 1');
+
+        // The rollback is real, and the survivor still carries no `errors`.
+        expect(res.results[1].errors).toBeUndefined();
+        expect(t.rows.has('t1')).toBe(true);
+        expect(t.rows.has('t2')).toBe(true);
+        expect(t.rows.has('t3')).toBe(true);
+    });
+
+    it('the NOT_ATTEMPTED message names record 2 too — it reads the same index', async () => {
+        // ⛔ Not assumed to ride along with the two 「unknown error」 sites. This
+        // wording never says 「failed」 and never quotes a cause, so its only
+        // defect was the INDEX — but that index is the shared one, so it moves
+        // with the locator. Pre-fix this read "atomic batch aborted by record 1".
+        const { res } = await atomicRunReachingBothInterpolations();
+
+        expect(res.results[3].errors[0].message).toBe('atomic batch aborted by record 2');
+    });
+
+    it('a rollback caused by a survivor ALONE reports no failure at all', async () => {
+        // Nothing threw: `runAtomicBatch` aborted on `outcome.failed > 0`, which
+        // a lone survivor satisfies. There is no causal ERROR to quote, so the
+        // message must not invent one — and must not call the survivor a
+        // failure either.
+        const t = makeCountingEngine('t2');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.batchData({
+            object: 'showcase_private_note',
+            request: {
+                operation: 'delete',
+                records: [{ id: 't1' }, { id: 't2' }, { id: 't3' }],
+                options: { atomic: true },
+            },
+        } as any);
+
+        expect(codesOf(res)).toEqual(['ROLLED_BACK', undefined, 'ROLLED_BACK']);
+        // No row in the whole response carries a fault entry of its own.
+        expect(res.results.some((r: any) => r.errors?.[0]?.code === 'RECORD_NOT_FOUND')).toBe(false);
+
+        // Pre-fix: "record 1 failed — unknown error".
+        for (const row of [res.results[0], res.results[2]]) {
+            expect(row.errors[0].message).toBe('record 1 did not succeed');
+            expect(row.errors[0].message).not.toContain('unknown error');
+            expect(row.errors[0].message).not.toContain('failed');
+        }
+        expect(t.rows.has('t2')).toBe(true);
+    });
+});
+
+describe('[#19452] CONTROLS — an ordinary batch, with no survivor in it, attributes exactly as before', () => {
+    const threeCreates = [
+        { data: { title: 'first valid' } },
+        { data: { title: POISON } },
+        { data: { title: 'third valid' } },
+    ];
+
+    it('batchData create: the stopped tail still names record 1 and quotes its error', async () => {
+        // The positive control for `expectAttributedTo`: this leg is green on
+        // the unfixed tree too, so a green above is about the survivor case and
+        // ⛔ not about an assertion that can no longer fail.
+        const t = makeCountingEngine('none_of_them');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.batchData({
+            object: 'showcase_private_note',
+            request: { operation: 'create', records: threeCreates },
+        } as any);
+
+        expect(codesOf(res)).toEqual([undefined, 'VALIDATION_FAILED', 'NOT_ATTEMPTED']);
+        expectAttributedTo(res.results[2].errors[0].message, res, 1);
+        expect(res.results[2].errors[0].message).toContain('title is invalid');
+    });
+
+    it('batchData create atomic: ROLLED_BACK still names record 1 and quotes its error', async () => {
+        const t = makeCountingEngine('none_of_them');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.batchData({
+            object: 'showcase_private_note',
+            request: { operation: 'create', records: threeCreates, options: { atomic: true } },
+        } as any);
+
+        expect(codesOf(res)).toEqual(['ROLLED_BACK', 'VALIDATION_FAILED', 'NOT_ATTEMPTED']);
+        expectAttributedTo(res.results[0].errors[0].message, res, 1);
+        expect(res.results[2].errors[0].message).toBe('atomic batch aborted by record 1');
+    });
+
+    it('updateManyData: the third bulk face cannot produce an errors-less non-success row', async () => {
+        // This face is covered by REASONING rather than by a survivor pin: it
+        // has no producer of a non-success row without `errors` — every push in
+        // `runUpdateManyLoop` is either `success: true` or a `toRowApiError`
+        // row — so `!success` and "carries a fault" still coincide on it. The
+        // reading is asserted, not asserted-about: every non-success row here
+        // carries an `errors` entry, which is exactly what the delete faces
+        // above violate.
+        const t = makeCountingEngine('none_of_them');
+        const p = new ObjectStackProtocolImplementation(t.engine);
+
+        const res: any = await p.updateManyData({
+            object: 'showcase_private_note',
+            records: [
+                { id: 't1', data: { title: 'renamed one' } },
+                { id: 't2', data: { title: POISON } },
+                { id: 't3', data: { title: 'renamed three' } },
+            ],
+        } as any);
+
+        expect(codesOf(res)).toEqual([undefined, 'VALIDATION_FAILED', 'NOT_ATTEMPTED']);
+        expect(res.results.filter((r: any) => r.success === false).every((r: any) => (r.errors?.length ?? 0) > 0)).toBe(true);
+        // And it shares the two builders, so the attribution moves with them.
+        expectAttributedTo(res.results[2].errors[0].message, res, 1);
+    });
+});
diff --git a/packages/metadata-protocol/src/protocol.ts b/packages/metadata-protocol/src/protocol.ts
index be9dd23ad9c..2204cd26953 100644
--- a/packages/metadata-protocol/src/protocol.ts
+++ b/packages/metadata-protocol/src/protocol.ts
@@ -12700,8 +12700,7 @@ export class ObjectStackProtocolImplementation implements
     ): BatchDataLoopOutcome {
         if (outcome.results.length >= records.length) return outcome;
 
-        const causeIndex = outcome.results.findIndex(r => !r.success);
-        const cause = causeIndex >= 0 ? outcome.results[causeIndex]?.errors?.[0]?.message : undefined;
+        const cause = this.locateBatchCause(outcome.results);
 
         const results = outcome.results.slice();
         for (let index = results.length; index < records.length; index++) {
@@ -12712,7 +12711,7 @@ export class ObjectStackProtocolImplementation implements
                 index,
                 errors: [{
                     code: 'NOT_ATTEMPTED' as const,
-                    message: `record ${causeIndex} failed — ${cause ?? 'unknown error'}; the batch stopped there. `
+                    message: (cause ? `${cause.clause}; the batch stopped there. ` : 'the batch stopped before this record. ')
                         + 'Set options.continueOnError to process the remaining records.',
                 }],
             });
@@ -12723,6 +12722,51 @@ export class ObjectStackProtocolImplementation implements
         return { results, succeeded: outcome.succeeded, failed: results.length - outcome.succeeded };
     }
 
+    /**
+     * The row that ENDED the run, plus the clause both builders interpolate.
+     *
+     * ⛔ NOT `findIndex(r => !r.success)`. `success` is the envelope's
+     * *outcome* bit, and its false arm is open by construction: it means "this
+     * row is not a success", which since #19412 also covers a row that MATCHED
+     * and was deliberately NOT removed — an outcome, not a fault, carrying no
+     * `errors[]` entry. Locating the cause with it named that survivor and then
+     * called the real error, sitting in the same array, "unknown" (#19452).
+     *
+     * The discriminator is the row's `errors[]` entry, which is the ONE per-row
+     * value whose declared meaning is a fault:
+     * `BatchOperationResultSchema.errors` is documented as *"Array of errors if
+     * operation failed"*, and the ADR-0087 v17 migration entry publishes
+     * `row.errors?.[0]?.message` / `.code` to consumers as exactly that read.
+     * It cannot widen the way the boolean did: every entry must carry an
+     * `ApiError.code` drawn from the closed `StandardErrorCode ∪
+     * ERROR_CODE_LEDGER` vocabulary, so an unregistered code fails
+     * `BatchOperationResultSchema.parse` — giving a non-fault ending an
+     * `errors[]` entry is a ledger widening in `packages/spec`, which is
+     * precisely the step both survivor sites declined to take. `message` is
+     * REQUIRED on `ApiErrorSchema`, so a located cause always has one and the
+     * 「unknown error」 fallback is gone rather than merely unreached.
+     *
+     * Scanned from the END because a run ends AT the row it stops on: every
+     * stop is a `break` in a loop's `catch`, immediately after that row was
+     * pushed. A fault that does NOT stop the run (the `Unknown operation:` arm
+     * records one and keeps going) must not be able to shadow the row that did.
+     *
+     * When no row recorded a fault at all the batch still ended for a reason —
+     * `runAtomicBatch` aborts on `outcome.failed > 0`, which a lone survivor
+     * satisfies — so the non-success row is named as what it is, ⛔ never as a
+     * failure and ⛔ never as an 「unknown error」.
+     */
+    private locateBatchCause(
+        rows: ReadonlyArray<BatchDataRowResult>,
+    ): { index: number; clause: string } | undefined {
+        for (let index = rows.length - 1; index >= 0; index--) {
+            const fault = rows[index]?.errors?.[0];
+            if (fault) return { index, clause: `record ${index} failed — ${fault.message}` };
+        }
+        const stalled = rows.findIndex(r => !r.success);
+        return stalled >= 0 ? { index: stalled, clause: `record ${stalled} did not succeed` } : undefined;
+    }
+
     /** The ordinary (committed) batch response — every row reports what it did. */
     private buildBatchDataResponse(
         operation: BatchUpdateRequest['operation'],
@@ -12771,21 +12815,25 @@ export class ObjectStackProtocolImplementation implements
         outcome: BatchDataLoopOutcome,
     ): BatchUpdateResponse {
         const attempted = outcome.results;
-        const causeIndex = attempted.findIndex(r => !r.success);
-        const cause = causeIndex >= 0 ? attempted[causeIndex]?.errors?.[0]?.message : undefined;
+        // [#19452] Same locator as the stopped-batch arm, for the same reason:
+        // `!success` stopped meaning "this row failed" when #19412 widened it.
+        // Both this builder's interpolations of the causal index read it, so
+        // neither the ROLLED_BACK message nor the NOT_ATTEMPTED one can name a
+        // row that merely survived — the two could not be fixed apart.
+        const cause = this.locateBatchCause(attempted);
 
         const results: BatchDataRowResult[] = records.map((record, index) => {
             const attempt = attempted[index];
             if (!attempt) {
                 return {
                     id: record.id, success: false, index,
-                    errors: [{ code: 'NOT_ATTEMPTED' as const, message: `atomic batch aborted by record ${causeIndex}` }],
+                    errors: [{ code: 'NOT_ATTEMPTED' as const, message: cause ? `atomic batch aborted by record ${cause.index}` : 'atomic batch aborted' }],
                 };
             }
             if (attempt.success) {
                 return {
                     id: attempt.id ?? record.id, success: false, index,
-                    errors: [{ code: 'ROLLED_BACK' as const, message: `record ${causeIndex} failed — ${cause ?? 'unknown error'}` }],
+                    errors: [{ code: 'ROLLED_BACK' as const, message: cause ? cause.clause : 'the atomic batch rolled back' }],
                 };
             }
             return { id: attempt.id ?? record.id, success: false, index, errors: attempt.errors };
diff --git a/scripts/engine-double-contract.pinned.json b/scripts/engine-double-contract.pinned.json
index 3a63859e226..f44d5c7c242 100644
--- a/scripts/engine-double-contract.pinned.json
+++ b/scripts/engine-double-contract.pinned.json
@@ -266,6 +266,21 @@
       "verb": "update",
       "pinned": 2
     },
+    {
+      "file": "packages/metadata-protocol/src/protocol.batch-causal-row.test.ts",
+      "verb": "delete",
+      "pinned": 1
+    },
+    {
+      "file": "packages/metadata-protocol/src/protocol.batch-causal-row.test.ts",
+      "verb": "findOne",
+      "pinned": 2
+    },
+    {
+      "file": "packages/metadata-protocol/src/protocol.batch-causal-row.test.ts",
+      "verb": "update",
+      "pinned": 2
+    },
     {
       "file": "packages/metadata-protocol/src/protocol.batch-not-attempted.test.ts",
       "verb": "delete",