Eight review deferrals from the last 3 weeks would benefit from tickets (22 of 30 were tracked)

Audit of every PR merged to `main` between 2026-07-31 and 2026-08-21 (617 PRs), checking whether
deferrals made in review — "follow-up PR", "separate issue", "out of scope here" — ended up with an
artifact somewhere.

**Most of them did.** 22 of the 30 real deferrals in that window are tracked: a GitHub issue, a
Linear ticket, a merged follow-up PR, or code on `main` that now does the thing. Deferring is
usually the right call and this is not a complaint about any of them. The eight below are the ones
that would benefit from a ticket, because today they exist only in a PR thread that is resolved or
scrolled past.

Two mechanical routes worked every time and are worth preferring over prose:
`@coderabbitai Make an Issue to do this in a followup PR on a future date` (6/6) and
`@Glary-Bot file … as follow-up issues in linear` (2/2, produced FE-1551 and FE-1552).

### The eight

**Item 1 has been moved out of this issue.** It is a CI-configuration hardening item in two workflows, and the specifics are being handled privately with the maintainers rather than described here. Nothing about it is urgent for a normal contributor; it is not reachable without a repo-member action. It remains on the unkept list and is being tracked separately.

**2. #14687 — `@coderabbitai tag this as a followup` produces no artifact.**
[Thread](https://github.com/Comfy-Org/ComfyUI_frontend/pull/14687#discussion_r3742160848). CodeRabbit
replied "Would you like me to create a GitHub issue for the follow-up?", nobody answered, and the
thread was resolved. The lint rule to prohibit the real-network string in `vitest.setup.ts` does not
exist and nothing references #14687. Worth knowing as a repo-wide fact: `tag this as a followup` is
not the same command as `Make an Issue`. *Owner: @christian-byrne, finding by @DrJKL.*

**3. #15221 — "Will followup on the nonblocking items."**
[Review](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15221#pullrequestreview-4932392345),
items enumerated
[here](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15221#issuecomment-5287460531). Both are
still live: `scripts/nodeDefLocaleSerializer.ts:105` annotates an `Object.fromEntries` call that
resolves to the `any` overload, so the annotation checks nothing — which undercuts that PR's own
stated goal; and `scripts/vite-define-shim.ts:32` uses `Object.assign` where the removed
`@ts-expect-error` at least would have reported itself unnecessary. *Owner: @benceruleanlu.*

**4. #15019 — two `SHOULD FIX (follow-up)` findings, threads still unresolved.**
[Batch abort](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15019#discussion_r3797214559) and
[post-`onRemoved` throw](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15019#discussion_r3797214567).
`src/platform/nodeReplacement/useNodeReplacement.ts:292` is still a bare `replaceWithMapping(...)`
inside the `for (const node of placeholders)` loop. The PR correctly made the guards throw for
visibility, but without a per-node catch one refused node unwinds the batch and permanently blocks
every later node. *Owner: @DrJKL, finding by @benjcooley.*

**5. #13880 — `translation-reviewers` team, "I'll follow up separately."**
[Review](https://github.com/Comfy-Org/ComfyUI_frontend/pull/13880#pullrequestreview-4769182672).
`CODEOWNERS:82-90` now carries a commented-out "Pending (needs a write grant before adding,
otherwise the line is inert)" block naming seven languages and their volunteer reviewers — `it`,
`uk`, `he`, `fa`, `tr`, `ar`, `pt-BR`. That record lives only in a source comment, so no query
anyone runs will surface it, and seven people who offered to review their own language are not
routed. *Owner: @christian-byrne.*

**6. #15062 — two changes described as written and validated locally, never pushed.**
[Bisect on `finish_reason: "length"`](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15062#discussion_r3772306012)
and [SDK-derived `ReasoningEffort`](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15062#discussion_r3772306226).
Most of that PR's follow-up did land (#15137, #15170 — openai SDK adopted, `fetch-depth: 0` +
`filter: blob:none` on both `ci-lint-format*` workflows). These two did not:
`scripts/i18n/translate.ts:146-148` still throws on truncation instead of bisecting the batch, and
`ReasoningEffort` is still a local union in `./config`. This shape is worth naming — "already fixed
locally" reads as done to everyone else in the thread. *Owner: @benceruleanlu.*

**7. #15063 — half of "doing the suggestions in a follow up" landed.**
[Comment](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15063#issuecomment-5274964930). The
single-reporter helper became #15346. The other suggestion —
[emit a Datadog warning for unmatched routes so we can monitor spikes](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15063#discussion_r3770774243)
— did not. `src/router.ts:90` is still a silent `{ path: '/:pathMatch(.*)*', redirect: '/' }`, so the
IR-105 failure class that PR was written for still produces no signal. *Owner: @christian-byrne,
suggestion by @huang47.*

**8. #15224 — members panel loading state.**
[Review](https://github.com/Comfy-Org/ComfyUI_frontend/pull/15224#pullrequestreview-4932463301):
*"Atomic update is the right call; a skeleton/loading state is separate follow-up work."* The atomic
multi-page fetch means a large workspace now shows an empty list for the duration of all pages
instead of the first 20 immediately. `MembersPanelContent.vue` has no loading state on `main`. #15235
is open but covers fixed height and scroll, not this. *Owner: @huntcsg, finding by @benjcooley.*

### Suggested handling

Items 1 and 4 are the two with user- or repo-visible consequences today; the rest are debt. Item 1
should probably move to a private advisory rather than stay here. Items 2, 5 and 6 are process
observations as much as work items — they describe three different ways a deferral evaporates
(a bot command that only looks like it files, a record kept in a source comment, and work described
as already done locally).

Full audit including the 22 kept deferrals and the verification for each:
`research/crossrepo/s04-followup-audit.md` in the ecs-migration workspace.

---

## Appendix: the 22 kept deferrals and their verification (appended 2026-08-23)

The full audit file referenced above lives in a review workspace repo readers cannot access, so the
kept-deferral verification is reproduced here. (The unkept-item detail is already enumerated above;
item 1 remains tracked privately, as stated.)

**Method:** all 617 PRs merged 2026-07-31 to 2026-08-21 pulled via GraphQL (issue comments + review
bodies + review threads), grepped for deferral language, then each deferral chased to an artifact
(issue, PR, Linear ticket) or to code on `origin/main` at `28bd73af29`. Run date 2026-08-21.

| | Count |
| --- | ---: |
| Merged PRs scanned | 617 |
| PRs with deferral language | 121 |
| Comments matching deferral language | 326 |
| Deferrals with a stated intent to follow up (after triage) | 30 |
| Kept - artifact found (issue / PR / Linear) or code now does the thing | 22 |
| Unkept - no artifact, code unchanged | 8 |

### Kept deferrals, each verified

| PR | Artifact |
| --- | --- |
| #15029 | #15215, #15216, #15217, #15227 (CodeRabbit) + FE-1557 |
| #14556 | #14737 |
| #14555 | #14736 |
| #14460 | #14641, #14465 |
| #15033 | FE-1551, FE-1552 (Glary-Bot, as instructed in-thread) |
| #14615 | #14645 |
| #14613 | #14738 + per-item issues, each with an owner |
| #14541 | #14629, #14630, #14631 |
| #14765 | #15220 (five items) |
| #14945 | Comfy-Org/cloud#6801 + siblings |
| #15311 | FE-1681 |
| #13783 | BE-3447 |
| #15125 | BE-7189 |
| #14836 | #15011 merged (Vitest cleanup oxlint rules) + #15308, #15309, #15370 |
| #14347 | #14951 merged (lint: ban new hand-written Zod schemas for remote API types) |
| #14574 | #14789 merged (search highlights rendered as text) |
| #14612 | code now does it - `throwOnTimeout: true` at `WorkspaceAuthGate.vue:128,260` |
| #14041 | #14386 |
| #14133 | #14480 superseded and split into #15001/#15009/#15017/#15018, all merged; the `test.fails` markers it deferred are gone from `main` |
| #15063 | #15346 (helper half; the Datadog-warning half is item 7 above) |
| #15062 | #15137, #15170 (majority of the follow-up; the two remaining gaps are item 6 above) |
| #14352 | accessibility + fixture-desync follow-ups taken; the remaining one is item 1 above, tracked privately |

### Notes for whoever runs this next

- Two mechanical routes are reliable and should be preferred over prose: `@coderabbitai Make an Issue to do this in a followup PR on a future date` (6/6) and `@Glary-Bot file ... as follow-up issues in linear` (2/2).
- `@coderabbitai tag this as a followup` is **not** one of them. It produces a thread reply and a question ("Would you like me to create a GitHub issue?") that nobody answers, and then the thread gets resolved. #14687 is the clean instance.
- "I have it fixed locally, not pushed yet" (#15062) is a distinct and higher-risk shape than "let's do this later": it reads as done to everyone else in the thread.
- A deferral recorded only in a source comment (#13880's `CODEOWNERS` block) is invisible to every query anyone will run.

Start with the eight unkept items in this issue and the referenced review threads, then inspect the named files such as scripts/nodeDefLocaleSerializer.ts, useNodeReplacement.ts, CODEOWNERS, scripts/i18n/translate.ts, src/router.ts, and MembersPanelContent.vue. Done means each reachable deferral has a separate tracked artifact or a verified resolution; item 1 is explicitly private and not reachable to normal contributors.

Written by the indexing model from the issue text.