test: pin keptIds branch and parent-container observation in dangling-link scrub characterization

[christian-byrne]

Test-only follow-up to #162 (benceruleanlu's approve-with-comments, 2026-09-02T21:26:01Z). Both non-blocking items addressed; no production logic changes.

Full context for agent readers

benceruleanlu's approval on #162 listed two non-blocking suggestions:

1. "Exercise live-link preservation through an operation that actually calls scrubDanglingLinkRefs. Both delete fixtures currently remove every link, while the surviving-link case uses set_widget; a delete fixture with one removed link and one unrelated live link would pin the keptIds branch under KA-4."
2. "Observe the parent input/output containers, or use observeDeep, when asserting that set_widget emits no link-reference writes. Observing only the existing child Y.Array misses an equivalent whole-value replacement of port.links."

This PR does both:

• New case keeps live-link references when delete_node removes only the deleted node's own links: a four-node workflow where delete_node(20) removes link 1 while unrelated link 3 (40→50) stays live. Asserts the dangling endpoint refs on node 10 are scrubbed, node 20 is gone, and the live link's endpoint refs (outputs[0].links on 40, inputs[0].link on 50) survive — the keptIds branches of scrubDanglingLinkRefs.
• The no-link-reference-writes case now observes the parent inputs/outputs containers with observeDeep instead of only the child links Y.Array, so a whole-value replacement of port.links (or slot.link) is counted. A harness-sensitivity check at the end of the case performs exactly that whole-value replacement directly on the doc and asserts the observer fires — with the old child-array observer this assertion fails, proving the gap is closed, not vacuous.

Note: this repo is retired read-only (its own AGENTS.md; #168) with active development moved to ComfyUI_frontend/packages/comfy-multi-player via FE #16644 (still open). The follow-up is opened here following the post-retirement precedent of #172 (test-only, off standalone main) so the review items have a carrier before the workspace move lands; it should port (or be re-targeted) with the move.

Origin row: act-cmp-162-followup (awc-audit second-pass CONFIRMED_GAP). QUIET WEEK: assigned christian-byrne only; the approver is recorded here as the originating reviewer, not assigned or mentioned.

Evidence

Node v25.9.0 (nvm use 25), head db3e005cf7317eddb877c7fe8c113e8d6903ce82:

$ npx vitest run test/scrub-dangling-link-refs.test.ts
 Test Files  1 passed (1)
      Tests  5 passed (5)

$ npm run typecheck            # build + tsconfig.test.json + test/types
  (clean, exit 0)
$ npm run lint                 # 1662 warnings, 0 errors (pre-existing warnings)
$ npm run check:purity         # purity check PASSED
$ npm run check:pins           # pass (offline mode)
$ npm run check:profile-claims # pass
$ npm run check:coderabbit     # PASSED (5 block(s) from 4 profile(s))
$ npm run verify:corpus        # PASSED (8 files)
$ npm run check:imports        # INCONCLUSIVE — pre-existing: dependency-cruiser
                               # JSON unparsable under Node 25.9.0, same as #162
$ npm test                     # 968 passed / 8 failed — ALL 8 failures are
                               # test/check-import-graph.test.ts, verified
                               # pre-existing environmental (dependency-cruiser),
                               # same 8 documented on #162; untouched here

Harness-sensitivity proof (item 2): reverting only the observeDeep change back to the child-array observer makes the new sensitivity assertion FAIL (adds no encoded link-reference writes… ×), confirming the new harness catches the whole-value port.links replacement the old one missed.

[coderabbitai]

Warning

Review limit reached

Next included review available in 44 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available. Your 129 included PR review attempts over the past 7 days set your current allowance at 1 review per hour.

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: a3f3da4c-c10c-49c9-a069-62450553693e

📥 Commits

Reviewing files that changed from the base of the PR and between 84030a4 and db3e005.

📒 Files selected for processing (1)

• test/scrub-dangling-link-refs.test.ts

---

_{Comment @coderabbitai help to get the list of available commands.}