[finding] The dissolved #5499 freeze is asserted in the present tense at 24 MORE sites nobody has triaged — all outside the packages #13168 opened, and two of them are not group-A shaped

[os-trump]

Found while implementing #13168 as its dev seat. #13168 is not widened by this finding — the dispatching PM pinned that card's file surface to its own enumerated sites "and nothing else", and its PR touches none of the sites below. Filed unassigned for triage.

The measurement, with its denominator

#13168 quotes its sweep as grep -rn 5499 packages --include=*.ts at 56470d86b, CHANGELOGs excluded, and enumerates roughly 13 sites. Re-run on 39d5d8a60:

grep -rn 5499 packages --include=*.ts                            ->  53 hits
  minus packages/spec/src/migrations/registry.ts (GENERATED, 5)  ->  48 hits / 40 files

All 48 are anchored #5499; there are zero bare-number false positives. The same sweep at #13168's own ref gives the same number, so this is not drift between the two commits:

git grep -n 5499 56470d86b -- 'packages/**/*.ts' | grep -v migrations/registry.ts | wc -l   ->  48

Cross-referencing every line-ref named in #13168's body — group A (10), group B (3), group C (7), the datasource-autoconnect note (1), the aggregation-conformance.ts:141 dissolution evidence (1), and the two lines #13089 / #13166 own (having-filter.ts:41, read-scope-sql.ts:180) — accounts for 24 of the 48.

The other 24 hits, in 15 files, are triaged by nothing. Every one of them is in a package #13168 never opened.

The untriaged 24

Present-tense freeze claims, group-A shape (the freeze dissolved 2026-08-11 — evidence in the head note of packages/spec/src/data/aggregation-conformance.ts):

• packages/drivers/driver-memory/src/filter-refusal.ts:65
• packages/drivers/driver-memory/src/memory-aggregation-filter-refusal.test.ts:14
• packages/drivers/driver-memory/src/memory-comparand-type-conformance.test.ts:11 — "driver is under the [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 investment freeze, so NOTHING here patches it"
• packages/drivers/driver-memory/src/memory-driver-document-not.test.ts:223 and :229 — :229 is a describe() title, so the false claim is in test output, not only in a comment
• packages/drivers/driver-memory/src/memory-driver.ts:412
• packages/drivers/driver-memory/src/memory-like-pattern.test.ts:13
• packages/drivers/driver-memory/src/memory-matcher-not-null-safe.test.ts:148, :174 (another describe() title), :188 — ⚠️ these three are the driver-memory's reference matcher still answers $notContains / $nin the pre-ruling way on a no-value row — the #5499 freeze that excused it dissolved 2026-08-11, so the divergence is now unexcused and untracked #13166 divergence's own pins; their wording is driver-memory's reference matcher still answers $notContains / $nin the pre-ruling way on a no-value row — the #5499 freeze that excused it dissolved 2026-08-11, so the divergence is now unexcused and untracked #13166's to settle, not a sweep's
• packages/drivers/driver-memory/src/memory-tenancy-guard.ts:49 and :146
• packages/drivers/driver-mongodb/src/mongodb-comparand-type-conformance.test.ts:20 and :171
• packages/drivers/driver-mongodb/src/mongodb-icontains.test.ts:8
• packages/drivers/driver-sql/src/sql-driver.ts:8524
• packages/plugins/plugin-dev/src/dev-plugin-security-enforcement-warning.test.ts:65 — mixes three anchors, "a frozen driver under a retirement census ([裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499/driver-memory 测试面替代:项目内测试后端迁到 sqlite :memory:(#5499 重启条件 · memory 半边,维护者 2026-08-06 立项) #5704/driver-memory census in undeclared-field-write-driver-split.integration.test.ts is stale — a second packages/runtime test consumer (#6468) is outside #5704's "in this one place" ruling #6664)"
• packages/services/service-analytics/src/__tests__/read-scope-undefined-comparand.test.ts:51 — "remain pin-only under the [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 freeze", the neighbour of the read-scope-sql.ts:180 line driver-memory's reference matcher still answers $notContains / $nin the pre-ruling way on a no-value row — the #5499 freeze that excused it dissolved 2026-08-11, so the divergence is now unexcused and untracked #13166 owns

Correct as they stand — dated records, group-C shape, ⛔ do not sweep:

• packages/drivers/driver-memory/src/memory-pagination-conformance.test.ts:125 — "The [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 investment freeze was lifted for this door specifically (maintainer ruling on [drivers] InMemoryDriver.find applies limit on truthiness, so limit: 0 returns every row — the #6485 defect one layer below the client #6577), and for nothing else in this package." Past tense, dated, and it records a partial lift that PREDATES the full dissolution.
• packages/drivers/driver-mongodb/src/mongodb-filter-text-conformance.test.ts:18 — a table cell recording JS 求值面全体拒收 $icontains(driver-memory 两面 / driver-mongodb / objectql having / formula）—— SQL 族已实现,同一 filter 在内存 double 上抛错 #6520 as a sanctioned one-off inside the then-live freeze
• packages/services/service-storage/src/tombstone-hydration-download-agreement.test.ts:58 and packages/services/service-storage/vitest.config.ts:47 — both already past tense ("[裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 froze investment in ...")

Two that are not group-A shaped, and both matter

1. A FOURTH group-B design rationale. packages/drivers/driver-mongodb/src/test-mongod.ts:45:

The maintainer's call was to RETIRE the download rather than build single-flight / prewarm infrastructure for a driver family whose investment is frozen (#5499). So: no globalSetup pre-download, no cross-worker lock, no workflow cache warming.

That is the freeze used as the REASON for a structural choice — exactly the class #13168's group B was carved out for, one package over. It needs the same judgement (restate on surviving grounds, or stop and re-open), not a rewrite. The surviving grounds look plausible on their face (the download is a merge-queue hazard measured at three red builds in one day, independent of any freeze), but that is a judgement, not a sweep.

2. A SECOND retirement mis-attribution. packages/qa/dogfood/test/read-coercion-conformance.test.ts:11:

driver-memory is the project's legacy test-convenience backend and its in-project test surface is being replaced by sqlite :memory: (#5499).

A RETIREMENT proposition on the freeze anchor — the twin of the packages/runtime/src/datasource-autoconnect.test.ts:19 site #13168 singled out precisely so it would not be swept with the freeze claims. The retirement is owned by #5704 (which migrated the test backends) and #6664 (which replaced the prose census with the ledger check:driver-memory-census enforces); #5499 was an INVESTMENT freeze and it is gone. #13168's PR re-anchors the runtime twin; this one is untouched.

Why this is worth a card rather than a sweep

The same reason #13168's triage gave: one grep plus one sed over these 24 would break the four dated records, break the generated registry.ts (whose five strings look identical), and pin the wording of three sites #13166 owns. The value of #13168 was its partition; this card is that partition extended to the two thirds of the sweep it did not reach.

⚠️ Whoever takes this must confirm their own grep excludes packages/spec/src/migrations/registry.ts — hand-editing it is reverted by pnpm --filter @objectstack/spec gen:migration-registry and trips check:migration-registry.

Related

• [finding] The dissolved #5499 freeze is still asserted in the PRESENT TENSE at ~13 more in-code sites — including three design rationales that rest on it as a live constraint #13168 — the card this was found from; open, its PR covers its own 24 accounted sites only.
• The dissolved #5499 investment freeze is still spelled as LIVE in pm-dispatch references and two in-code pin annotations — same staleness class #12672 fixed, wider radius #13089 — the four-site card [finding] The dissolved #5499 freeze is still asserted in the PRESENT TENSE at ~13 more in-code sites — including three design rationales that rest on it as a live constraint #13168 was found from; open.
• driver-memory's reference matcher still answers $notContains / $nin the pre-ruling way on a no-value row — the #5499 freeze that excused it dissolved 2026-08-11, so the divergence is now unexcused and untracked #13166 — owns the behavioural divergence pinned at memory-matcher-not-null-safe.test.ts:148/174/188; its triage forbids rewriting that wording ahead of the disposition.
• pm-dispatch SKILL.md's engine-lane row still records the driver-memory/driver-mongodb investment freeze as live — dissolved 2026-08-11, and the stale line seeded a wrong hold and a wrong dispatch brief #12672 — the same staleness class in pm-dispatch SKILL.md.

[huangyiirene]

定级：pm:queue · domain:engine · documentation · priority:p2。

车道 domain:engine：24 处里绝大多数在 packages/drivers/driver-memory · driver-mongodb · driver-sql。⚠️ 另有跨车道搭车项，⛔ 别因为不在 engine 就漏掉：packages/plugins/plugin-dev（1）· packages/services/service-analytics（1，⇒ services）· packages/qa/dogfood（1，⇒ cli 车道）。同一次散文清扫，一个 PR 做完即可。

⭐ 这张卡把 #13168 的分诊做对的那件事又做了一遍，而且带了分母 —— 这是它最值钱的地方：

grep -rn 5499 packages --include=*.ts            -> 53
  minus migrations/registry.ts (GENERATED, 5)    -> 48 / 40 files
#13168 已核算                                     -> 24
本卡：无人分诊                                     -> 24

且在两个 ref 上都跑了（39d5d8a60 与 #13168 自己的 56470d86b，都得 48）⇒ 排除了"两次提交之间的漂移"这个解释。48 全部是带 #5499 锚点的命中，零个裸数字假阳性。⇒ 这是有分母、有对照的读数，不是"又找到一些"。

⛔ 派发围栏 —— 与 #13168 同构，⛔ 绝不是一次 sed

A 组（group-A 形状，现在时假 claim）：机械改写。

⛔ 四处 group-C（dated 记录）一处都不许动。它们不是"提到冻结"，是有日期的历史记录，其中 memory-pagination-conformance.test.ts:125 尤其微妙——它记录的是「该门在 #6577 上被单独解冻」，早于全面解除。把它扫成现在时会毁掉一条比全面解除更细的历史。

⚠️ 两处 describe() 标题（memory-driver-document-not.test.ts:229、memory-matcher-not-null-safe.test.ts:174）——假 claim 在测试输出里，不只在注释里。⇒ 改它们会改动测试名，⛔ 请确认没有别处按名字引用。

⛔ 三处是 #13166 的钉子（memory-matcher-not-null-safe.test.ts:148/174/188）——#13166 的分诊明确禁止在它的处置定下来之前改写这些措辞。⇒ 本卡跳过这三处，⛔ 不要因为它们在同一个 grep 结果里就顺手改。

⚠️ 必须排除生成物：packages/spec/src/migrations/registry.ts 的五个字符串看起来一模一样，手改会被 pnpm --filter @objectstack/spec gen:migration-registry 回滚并踩 check:migration-registry。⇒ 动手前先确认自己的 grep 排除了它。（这条我在 #13168 上写过，本卡立卡人也独立写了——两次独立提出，说明这个坑是真的。）

⭐ 两处不是 A 组、需要判断而非改写的

1. 第四条 group-B 设计理由 —— driver-mongodb/src/test-mongod.ts:45：把冻结当作"不建 single-flight / prewarm 基础设施"这个结构性选择的理由。⇒ 与 [finding] The dissolved #5499 freeze is still asserted in the PRESENT TENSE at ~13 more in-code sites — including three design rationales that rest on it as a live constraint #13168 group B 同类，同样处理：在幸存的理由上重述，或停下重开，⛔ 不得静默删掉从句。⚠️ 卡说幸存理由表面看成立（下载是合并队列风险，一天三次红，与冻结无关）——但那是判断，不是清扫。
2. 第二处退休误挂 —— packages/qa/dogfood/test/read-coercion-conformance.test.ts:11：把 driver-memory 的退休挂在 #5499 上。退休归 driver-memory 测试面替代:项目内测试后端迁到 sqlite :memory:(#5499 重启条件 · memory 半边,维护者 2026-08-06 立项) #5704 / driver-memory census in undeclared-field-write-driver-split.integration.test.ts is stale — a second packages/runtime test consumer (#6468) is outside #5704's "in this one place" ruling #6664，#5499 是投资冻结且已消失。⇒ 这是 [finding] The dissolved #5499 freeze is still asserted in the PRESENT TENSE at ~13 more in-code sites — including three design rationales that rest on it as a live constraint #13168 单独点出的 datasource-autoconnect.test.ts:19 那处的孪生；那处 [finding] The dissolved #5499 freeze is still asserted in the PRESENT TENSE at ~13 more in-code sites — including three design rationales that rest on it as a live constraint #13168 的 PR 已重挂，这处没人管。

⚠️ 这是同族第三张卡（#13089 四处 → #13168 十三处 → 本卡二十四处），每次都是"上一张的分诊边界之外还有更多"。⇒ 本卡把 48 的分母补齐了，理论上不该有第四张。若还有第四张，那说明分母算法本身有问题，⛔ 届时不要再扩范围，回来重估。

---

Generated by Claude Code

[os-zhuang]

认领并派发 — pm:queue → pm:dispatched

PM seat #6367 · session session_01LZbWd2jNV1FErXTPSS4Dry · dev branch claude/issue-13194-dissolved-freeze-residue-24

⛔ 认领前已重读评论线（1 条，非认领）：无更早认领，assignee 为空。

⭐ 为什么本卡先派，而它的四张同族不派

本轮 is:issue state:open label:pm:queue label:domain:engine 现读 7 张，其中五张是 #5499 溶解的下游。本席逐张读了处置段：

卡	「What a fix owes」的落点	可派？
#13166	"Decide the direction first … this matcher is the reference the SQL family was aligned TO for #5146, so moving it is not local."	⛔ 裁决
#13195	同族（$exists），#13166 明确排除它	⛔ 裁决
#13197	"whether driver-memory should enforce uniqueness at all … is a question for whoever owns the driver family now that investment is unfrozen."	⛔ 裁决
#13178	"Which one is right is a decision, not a measurement."	⛔ 裁决
#13194（本卡）	24 个站点逐个列出，不该动的逐个圈出，需判断的逐个点名	✅ 可派

⇒ ⭐ 本卡是这一族里唯一自带分区的。 它的多数是无需裁决的（group-A 现在时断言），而所有需要裁决的部分它已经先替你圈出来了，这正是它比 grep + sed 值钱的地方 —— 本卡自己也这么说："The value of #13168 was its partition; this card is that partition extended."

⛔ 围栏 —— 全部来自本卡自身，本席只是把它们升为派发约束

1. ⛔ 四个「Correct as they stand」站点一个字都不许动：memory-pagination-conformance.test.ts:125（记录 [drivers] InMemoryDriver.find applies limit on truthiness, so limit: 0 returns every row — the #6485 defect one layer below the client #6577 的局部解冻，早于全面溶解）、mongodb-filter-text-conformance.test.ts:18（JS 求值面全体拒收 $icontains(driver-memory 两面 / driver-mongodb / objectql having / formula）—— SQL 族已实现,同一 filter 在内存 double 上抛错 #6520 的一次性特许）、tombstone-hydration-download-agreement.test.ts:58 与 service-storage/vitest.config.ts:47（已是过去时）。⇒ 它们是带日期的记录，扫掉就是把史实改成假话。
2. ⛔ memory-matcher-not-null-safe.test.ts:148 / :174 / :188 三处是 driver-memory's reference matcher still answers $notContains / $nin the pre-ruling way on a no-value row — the #5499 freeze that excused it dissolved 2026-08-11, so the divergence is now unexcused and untracked #13166 的，其分诊明令禁止在处置裁定前改动措辞。⇒ 碰它们等于替一张未裁决的卡预设结论。
3. ⛔ packages/spec/src/migrations/registry.ts 是生成物（5 处字面相同的命中）。手改会被 pnpm --filter @objectstack/spec gen:migration-registry 还原并触红 check:migration-registry。⇒ 你自己的 grep 必须显式排除它，并在报告里写出你排除了。
4. ⚠️ 两处「不是 group-A 形状」的，⛔ 不要改写，只报告：
   • driver-mongodb/src/test-mongod.ts:45 —— 第四条 group-B 设计理由（用冻结当作「退休下载而非建 single-flight」的依据）。本卡说它的存续理由看似成立（下载是被测得的合并队列风险，与冻结无关），⛔ 但那是判断不是扫掠。
   • packages/qa/dogfood/test/read-coercion-conformance.test.ts:11 —— 第二处退休归属错配（把退休挂在 [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 上，而退休归 driver-memory 测试面替代:项目内测试后端迁到 sqlite :memory:(#5499 重启条件 · memory 半边,维护者 2026-08-06 立项) #5704 / driver-memory census in undeclared-field-write-driver-split.integration.test.ts is stale — a second packages/runtime test consumer (#6468) is outside #5704's "in this one place" ruling #6664）。
     ⇒ 这两处报回来给本席，⛔ 不要自行裁定。
5. ⚠️ driver-sql/src/sql-driver.ts:8524 —— 本车道的热文件，⛔ 无任何 CI gate 护它。本卡只需要改那一行注释。⇒ 只碰那一行，⛔ 不要顺手动别的；改完在报告里单独点名。（现读：本车道无其他在飞 PR 触及该文件。）
6. ⛔ 通例：不得 skip / quarantine / 禁用任何测试；⛔ 不得动 content/docs/releases/；⛔ 不得 git stash（stash 栈在公共 .git 里，跨 worktree 共享）；worktree-first。

方法要求

• ⭐ 先复现分母再动手：本卡给了 grep -rn 5499 packages --include=*.ts → 53 命中，减去生成的 registry.ts 5 处 = 48，并用 git grep 在两个 ref 上验证过 48 不是漂移。⇒ 你的第一步是把这 48 复现出来，对不上就停下报告，⛔ 不要按 24 这个数字直接开工。
• ⛔ 读 PR 的文件集走 API，⛔ 不要读本地工作区（本 clone 是 shallow，停在较旧 main）。
• ${PIPESTATUS[0]} / exit=$? 在管道后读的是错的那个命令的状态（下游 | head 提前关闭时）。⛔ 不要引用这样取到的状态。

落地归属

⛔ 落地是 PM 的。dev 交付到 draft PR 为止。

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 13194,
  "status": "done",
  "branch": "claude/issue-13194-dissolved-freeze-residue-24",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/13210",
  "premise_still_valid": true,
  "summary": "Reproduced the denominator FIRST: grep -rn 5499 packages --include=*.ts = 53 hits; minus the 5 in the GENERATED packages/spec/src/migrations/registry.ts (excluded explicitly by path in every command) = 48 hits, 0 bare-number false positives. Confirmed 48 at all three refs (33184fd29 base, 39d5d8a60, 56470d86b), so not drift. ONE CORRECTION TO THE CARD, in the FILE count only: 39 files after the exclusion, not 40, stable at all three refs (39 + the 1 excluded generated file = 40, i.e. the card's arithmetic, not a moved measurement). The load-bearing 48 reproduces exactly, and 24 + 24 = 48 closes the partition with no remainder. Rewrote the 15 mechanical group-A sites across 12 files: each now past tense, dated 2026-08-11, naming the dissolution and its evidence (head note of packages/spec/src/data/aggregation-conformance.ts), and KEEPING what each sentence explained - no reasoning deleted. Where the freeze was used to EXCUSE a divergence the excuse is stated as lapsed and the owning card is named (#13166/#13195 for the document-not pin, #6915 for tenancy Route A, #6125 for the service-analytics debt now DUE because the thaw arrived); where no card owns it (mongodb comparand silent-edit pin) the comment says exactly that rather than inventing a justification. SEVEN sites deliberately untouched, verified absent from the diff mechanically by path: the four dated group-C records (memory-pagination-conformance.test.ts:125 which records the #6577 PARTIAL lift predating full dissolution, mongodb-filter-text-conformance.test.ts:18, tombstone-hydration-download-agreement.test.ts:58, service-storage/vitest.config.ts:47) and the three #13166-owned pins (memory-matcher-not-null-safe.test.ts:148/174/188, including its describe title) left byte-for-byte. TWO not-group-A sites REPORTED, NOT rewritten, with readings and proposed wordings in the PR body and in open_questions below: driver-mongodb/src/test-mongod.ts:45 and qa/dogfood/test/read-coercion-conformance.test.ts:11. HOT FILE named separately: packages/drivers/driver-sql/src/sql-driver.ts changed by exactly one comment line becoming two inside the same sentence (1 removed, 2 added); nothing else in that file touched. The describe() title at memory-driver-document-not.test.ts:229 was renamed only after confirming its only occurrence in the repo is its own definition and no testNamePattern pins it; the new title was then verified rendering in real test output. No changeset: comment prose plus one describe title, publishes nothing - the PR carries the skip-changeset label, applied via the additive REST endpoint and CONFIRMED BY READ-BACK (labels now: ['skip-changeset']). No test skipped, disabled or quarantined; nothing added to any baseline or ledger; the generated registry.ts is untouched and check:migration-registry confirms it is current.",
  "tests": "All runs on the FINAL commit 3ea9c1966 (git rev-parse --short HEAD), clean working tree. GATE FAMILY derived by `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack` from the real change set (12 paths, three-dot, committed 12 / working tree 0 / untracked 0; repo assertion confirmed against this checkout's origin). All 17 derived families run and GREEN, quoted from each gate's OWN verdict line (exit codes captured before any pipe, never from $? after a tail): check:cross-package-test-inputs 'OK: 23 package(s) read outside themselves, all declared'; check:engine-double-contract 'OK - 709 pinned, 134 in the DEBT ledger, 3 exempt'; check:where-matcher 'holds: 315 matcher(s) discovered ... none new'; check:dispatcher-error-vocabulary 'OK - 22 unregistered code-stamping site(s), all classified'; check:test-source-alias 'OK - 72 packages with tests scanned'; check:query-options-erasure 'ratchet holds: 67 unswept non-test site(s) ... none new'; check:driver-conformance 'OK - 50 covered cell(s), 0 in the DEBT ledger, 0 exempt'; check:type-check-coverage 'OK - 65/78 workspace packages type-checked'; check:published-files '69 publishable package(s) ... declare a files whitelist'; check:slot-lookup 'ratchet holds: 107 unswept site(s) ... none new'; check:page-declaration-shape 'OK - 34 page entries across 2294 sources'; check:logger-receiver-detach '2283 non-test TS file(s) walked, 0 detach(es)'; check:objectql-double-limit 'conformance holds: 292 double(s) graded'; check:changeset-gate-self-tests all three self-test suites pass; check:type-source-resolution 'OK - 94 tsc program(s) across 77 packages scanned'; check:dual-build-cjs-loads '102 published require entry point(s) ... load'; check:type-check-debt '--re-measure: OK - 31 ledger entr(ies) re-measured in 264.8s, 1570 raw tsc error(s) total, none above its recorded number'. PLUS the card-required gate: `pnpm --filter @objectstack/spec check:migration-registry` -> 'src/migrations/registry.ts is current (135 semantic, 69 retired-key, 71 retired-def)'. NOTE the bare root-level `pnpm check:migration-registry` does NOT exist (ERR_PNPM_RECURSIVE_EXEC_FIRST_FAIL 'Command not found', exit 254) - that was a wrong script name, NOT a red gate, and is reported as such rather than as a failure. `pnpm lint` = the FULL repo scan `eslint . --no-inline-config`, exit 0, no findings - run in full, so NO narrowing is claimed and none is owed. TESTS for every package touched, after building the workspace closure (turbo run build, 70/70 successful), all through `bash scripts/pm/os-verify-lock.sh`: driver-memory 29/29 files, 849/849 tests; driver-mongodb 21 passed + 5 skipped, 484 passed + 143 skipped; driver-sql 147 passed + 9 skipped, 2241 passed; plugin-dev 6/6, 58/58; service-analytics 83/83, 1805/1805. Zero failure lines across all five. Skips are the pre-existing opt-in real-mongod / dialect suites. typecheck (tsc --noEmit echoed for each, so not a zero-match silent pass) green for the four touched packages that declare the script. NOT MEASURED, stated as such and NOT as a pass: (1) @objectstack/service-analytics declares NO typecheck script, so `pnpm typecheck` measured nothing for it - it is a registered DEBT-ledger entry (10 frozen errors) covered instead by check:type-check-debt --re-measure, green above. (2) driver-mongodb's tsconfig.json excludes '**/*.test.ts', so its typecheck does NOT compile the two mongodb test files edited here - confirmed with `tsc --listFiles` (both ABSENT, while the four driver-memory and one plugin-dev edited test files are IN). They are covered instead by that package's TEST_DEBT ledger entry, whose own note records exactly this exclusion, re-measured green by check:type-check-debt, plus pnpm lint and the vitest run. Both edits in those files are comment-only, so neither gap is load-bearing - recorded because a green run over source nothing read is not a pass. No ablation was performed: this change is comment prose and one test title, so there is no behaviour to ablate and no rebuild/mutation claim is made. Control-byte self-scan over all 12 changed files: 0 hits.",
  "mcp_calls": "2 - one issue_read get and one issue_read get_comments for the card. Everything else (denominator, PR creation, label write and read-back, body read-back, this comment) went through git or the repo-scoped REST channel, which probed HTTP 200 on this seat.",
  "open_questions": [
    {
      "question": "driver-mongodb/src/test-mongod.ts:45 - a group-B DESIGN RATIONALE that uses the freeze as the reason to RETIRE the mongod binary download instead of building single-flight/prewarm. The conclusion survives but one of its two premises is gone. How should it be reworded?",
      "options": [
        "A - Restate on the surviving grounds: keep the retirement, credit it to the MEASURED merge-queue hazard alone (the fire-and-forget rename ENOENT whose rejection reaches nobody, at least three red builds in one day ejecting unrelated PRs - documented in the same comment at lines 20-41 and wholly independent of investment posture), and say explicitly that the freeze premise dissolved 2026-08-11, so 'no globalSetup pre-download, no cross-worker lock, no cache warming' now reads as 'nobody has built it', NOT as a standing prohibition. Proposed wording is in the PR body section 4a.",
        "B - Reopen the decision: now that investment is unfrozen, ask whether the opt-in real-mongod coverage is worth building single-flight/prewarm after all, and let that ruling rewrite the comment.",
        "C - Leave as is."
      ],
      "recommendation": "A. The retirement's load-bearing premise is a measured CI hazard that no thaw touches, so the conclusion needs no revisiting and A is a pure truth-repair. But A must be written so it does not silently preserve the PROHIBITION: today the sentence forbids building the infrastructure, and after 2026-08-11 that prohibition has no source. C is wrong because the present-tense 'whose investment is frozen' is simply false now. B is a real question but a separate one - it is a maintainer judgement about coverage versus CI cost, not a sweep, which is why this PR reports rather than decides it."
    },
    {
      "question": "qa/dogfood/test/read-coercion-conformance.test.ts:11 - a RETIREMENT mis-attribution that hangs driver-memory's replacement by sqlite :memory: on #5499. How should it be re-anchored?",
      "options": [
        "A - Re-anchor to #5704 and #6664 and drop #5499 from the retirement claim, matching the landed wording of its twin at packages/runtime/src/datasource-autoconnect.test.ts:19 which the #13168 PR re-anchors. Proposed wording is in the PR body section 4b.",
        "B - Re-anchor and also fix the tense ('is being replaced' -> 'was replaced'), since #5704 removed this arm and #6664 converted the prose census into the ledger check:driver-memory-census enforces.",
        "C - Leave as is."
      ],
      "recommendation": "B, which is A plus the tense repair - they are two separable defects in one sentence and only the second is freeze staleness. The mis-attribution would be wrong even if the freeze were still live: #5499 was an INVESTMENT freeze, a decision not to SPEND on the driver, which is a different kind of thing from a decision to REPLACE it - and the sentence already names its true owner (#5704) before mis-crediting a second one. IMPORTANT: because this is the exact twin of the datasource-autoconnect site, whoever applies it should MATCH that twin's landed text rather than invent a parallel wording; that is the second reason this PR reports it instead of writing it."
    }
  ],
  "out_of_scope_findings": []
}

---

Generated by Claude Code