[lilyshen0722]

What. docs/development/review-checklist.md rule 24 gains a second earned instance: the same silent-drop defect in its update shape, which the rule did not previously carry.

Why this instance and this rule. vera asked at 74186 where #1898's strict-subdocument finding belongs, and suggested rule 37's family. The class is rule 24's: a strict schema silently drops an undeclared path on the write, so a fixture that builds its own object proves the reader can read a field and never that the field can be there. Rule 37 answers a read-side question — whether a read tells you the KEY exists, the VALUE is non-empty, or the value can AUTHENTICATE. Filing a write-side drop under the read-side rule would hide it from the next reviewer grepping for it, so it goes where the class already lives.

The instance. In #1898 the writer was an updateOne $set rather than a constructor: $set { 'config.webhookUrlRef': ref } on a row the Discord connect route had just created was dropped in silence because Integration.config is a strict subdocument that did not declare the path. backend/models/Integration.ts now declares webhookUrlRef: String beside the other Discord config paths. Same tell, same instrument as #1734/TASK-043's constructor case: the route suite stayed green on inline-built fixtures, and the only assertion that saw the ref missing is against real Mongo after the create call (backend/__tests__/unit/routes/integrations.discordTokenCopy.test.js).

For the gate (@wren). Two things worth your scepticism rather than mine:

1. Is rule 24 the right home, or does the write-side drop deserve its own number? My argument is that the class is identical (strict schema + undeclared path + a green read-side test) and only the writer's shape differs; if you read it as two rules, say so and I will cut the second properly.
2. Is the instance accurate? The locators are checkable: the declaration in backend/models/Integration.ts, the real-Mongo assertion above, and the feat(discord): encrypt the webhook URL at rest, keyed by (integrationId, kind) (TASK-124 part 2) #1898 diff for the drop itself.

Docs-only, one file, no code. Rule 37 landed as #1894; this adds no number.

Resolved after the gate's hold (272ca93d → 8aa22c85). wren: rule 24 is the right home (no new number), the real-Mongo suite claim holds, and two locators did not hold — both cited paths contain the ref only on the unmerged #1898 (git grep webhookUrlRef origin/main -- backend frontend docs returns nothing), so the clause now names #1898 as the carrier of both, describes the witness by mechanism (create route against a MongoMemoryServer-backed model, read back with .lean()) and carries the measured census (5 suites name the ref; the other 4 mock Integration or seed the ref themselves). Third correction: my "grepping webhookUrlRef returned nothing" sentence described main before #1898 lands, not the rev the lesson came from. It now says the instructive thing instead: the name already appeared across the backend and the schema declaration was the one place it was missing, so a grep for the name cannot find this defect — the comparison rule 24's check opens with can.

Fourth correction (720e6d21 → this head), two swaps only: the count is dropped ("the name already appeared across the backend") and "this rule opens with" becomes "this rule's check opens with" — the rule opens with its thesis; the comparison is the first step of its check ("The check is two greps and one round-trip…"). Reverse-applying those two phrases to the new line reproduces 720e6d21 byte-for-byte, so nothing else moved.

Second hold round (8aa22c85 → this head). wren: the writer is not updateOne (it is findByIdAndUpdate on both Model writers, routes/integrations.ts:474 and discordService.ts:345; the migration's updateOne runs on mongoose.connection.collection() and strict mode never touches it), and config is a nested path of a strict schema, not a subdocument (schema.path('config') undefined, schema.nested.config true, schema.get('strict') true — measured from the model, and the wording rule 34 already uses). The clause now teaches the split — Model writes are stripped, raw-collection writes are not — so a grep for updateOne cannot lead a reader to the one writer that cannot drop. The backfill's drop is marked derived, not observed.

Merge order: after #1898 — this line cites code that only #1898 puts on main. Enqueue #1898 (cleared at 715e187c) first, then this at 8aa22c85.