automation: the region refusal leaves a deeper suspended run uncancelled when a middle link of the suspension chain cannot be read

domain:server#3 · os-steve · session_01ADHkmQJdgarUjPLNgFBY67

Filing gate: ① out-of-scope finding of the #29166 dev (os-dev-report 6576781813, open_questions[1]), confirmed by the seat reading resolveSuspensionChain on main (read, not run). Filed by domain:server seat 3. ⛔ Not a claim. Suggested-grade: p3 · same defect class as #29166 item (1), one link deeper.

Reach: a region whose body calls a subflow or map whose child run suspends through a chain of two or more linked runs, while the store answers for the child but fails for a deeper link.
Reader who acts: the domain:server lane (packages/services/service-automation/src/engine.ts resolveSuspensionChain, dag-traversal.ts runRegion).

What happens (source reading)

automation: the region refusal leaves its child run uncancelled with only a warn when the suspension chain cannot be read, and a region call reached through a child's top-level call still arms #29166 item (1) (PR fix(service-automation): the region refusal cancels its child run even when the suspension store cannot be read #29329) makes runRegion call cancelRun(childRunId, reason) once when resolveSuspensionChain(childRunId) comes back empty, so the strict loader owns the error record.
resolveSuspensionChain (engine.ts ~L6975) walks the chain with the degrading loadSuspendedRun, which turns a store read failure into null. When the child is readable but a deeper link's read fails, the walk stops early and returns a partial chain. The deeper run is then neither cancelled nor recorded at error; only the degrading read's warn remains.

Wanted

When runRegion cancels the chain, walk it with the strict loader (the shape cancelRun already uses), cancel every link it can read, and record at error each link it cannot read, naming the run id. The region refusal itself is unchanged and still thrown after.
Tests: a three-run chain whose middle link's store read fails; the readable links are cancelled; one error record names the unreadable link; ablation.
Clause-② expected: no (cancels that should have happened now happen). Metadata-protocol: none.

Dedupe: searched resolveSuspensionChain, region suspension chain cancel, partial chain deeper run not cancelled; nearest #29166 (empty chain only), #28897 (closed). New.

Links back: #29166, PR #29329.

triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U

Path: none

Triage: p3 · `domain:server` · `pm:queue` · 2026-11-02T13:02Z

⛔ Not a claim.

Read fresh. Read from source: when a middle link of the suspension chain cannot be read, the region refusal leaves the deeper suspended run uncancelled. This is automation: the region refusal leaves its child run uncancelled with only a warn when the suspension chain cannot be read, and a region call reached through a child's top-level call still arms #29166 item (1) one link deeper, p3. Lands after PR fix(service-automation): the region refusal cancels its child run even when the suspension store cannot be read #29329.

Generated by Claude Code

domain:server#3 · os-steve · session_01ADHkmQJdgarUjPLNgFBY67

Claim: PM loop round 1 · card thread · 2026-11-02T13:15Z
Session: session_01ADHkmQJdgarUjPLNgFBY67 (the seat); the card thread is session_01ADHkmS5997viAw9y7Jknnz
Account: os-steve (the seat's linked user; the card's assignee)
Branch: claude/project-thread-y9hvos
Worktree: the card thread's own container checkout
Domain: domain:server
Seat: domain:server#3
Ruling-ref: the maintainer, verbatim, in the project thread: 「开 server 2、3、5 号席位, 前任已经下班，接手所有开发，包括跟进之前未完成的开发落地。」
File surface: packages/services/service-automation/src/dag-traversal.ts (runRegion cancel walk) and engine.ts (a strict walk beside resolveSuspensionChain) + tests
Container & model: card thread cmsg_01FdL1XeUyvVNxLreh3F4Qi9JeMXnsbKRJxMbW9ZUq9nDf, mode:thread, model: opus (--tier names no path mandate: PM's call, default), effort: medium
Clause-②: no — cancels that should have happened now happen
Metadata-protocol: none
Responsibility: the card thread measures, codes, tests and opens the PR; the seat reviews and lands
Thread-read: 6577811795
Premise: holds (site re-read on origin/main 57cde8cb by the seat before dispatch)
Serial constraints cleared: PR #29329 (#29166) edits runRegion and is in the merge queue: the PR opens only after it merges, on a merged main
Resume-from: none — fresh from origin/main

State: pm:queue → pm:dispatched, assigned to os-steve, recorded by this claim. Order: project file server-seat3/standard-order.md + server-seat3/29333-delta.md.

domain:server#3 · os-steve · session_01ADHkmS5997viAw9y7Jknnz
os-dev-report

{
  "issue": 29333,
  "status": "done",
  "branch": "claude/project-thread-y9hvos",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/29403",
  "session": "session_01ADHkmS5997viAw9y7Jknnz",
  "premise_still_valid": true,
  "summary": "runRegion's region-refusal cancel now walks the child's suspension chain through resolveSuspensionChainForCancel: the same one walk as resolveSuspensionChain (now walkSuspensionChain with the loader as its only parameter) read through loadSuspendedRunStrict. Every read link is cancelled deepest first; the link whose read threw is still handed to cancelRun and one error record names it, since runs below it were never reached. resolveSuspensionChain keeps the degrading loader for the resume gate. #29329's empty-chain branch is subsumed (unreadable child = unreadable link); its two tests pass unedited.",
  "tests": "head a06cb531: service-automation typecheck exit 0; vitest Test Files 260/260, Tests 3342/3342 (os-verify-lock VERDICT command-exit 0); subflow-region-suspension.test.ts 24/24. New: middle-link-unreadable three-run chain + all-readable CONTROL. Ablation on e6d2696f via ablation-replace (loadSuspendedRunStrict -> loadSuspendedRun in the cancel walk): anchor 1->0, blob c937469c8834->ebabde87953c, 3 failed / 21 passed (new case + both #29166 cases), CONTROL green; restored blob == HEAD, git diff HEAD empty. src import, no dist leg.",
  "gates": "dispatch-gates --commands on a06cb531: 62 derived; --ran: 62 run, 0 NOT-MEASURED, all exit 0. check-plugin-teardown-shape --self-test and check:dual-build-cjs-loads first exit 3 (prerequisite), re-run exit 0 after git fetch of fixture 621a4876 and turbo build ./packages/** (78/78). pnpm lint exit 0 (same lock run, VERDICT command-exit 0). CI check-runs: in_progress at report time.",
  "dogfood_battery": "n/a — no ! in the PR title and Clause-②: no without a (narrowing) arm",
  "line_budget": "+144 / -18 over 4 files (changeset 10, dag-traversal.ts +30/-14, engine.ts +43/-4, test +61)",
  "files_changed": [
    ".changeset/29333-region-refusal-deeper-chain-cancel.md",
    "packages/services/service-automation/src/dag-traversal.ts",
    "packages/services/service-automation/src/engine.ts",
    "packages/services/service-automation/src/subflow-region-suspension.test.ts"
  ],
  "deviations": [
    "The pre-merge branch first merged #29329's head branch (a72981df) to build ahead of its landing; after #29329 squash-merged, merging main conflicted in the two files and was resolved to this branch's side. The PR was opened only after #29329 merged.",
    "No forced test edits."
  ],
  "mcp_calls": "4 — mcp__github__pull_request_read x4 (get_diff/get/get_status on #29329, get on #29403); no MCP writes",
  "api_writes": "2 — fleet relay pr_create (#29403), post-stamped os-dev-report comment on #29333",
  "open_questions": [
    {
      "question": "Runs BELOW the unreadable link stay parked: their ids live only in that link's correlation and SuspendedRun has no parent pointer. The error record names the unreadable run and tells the operator to cancel the rest once the store answers. Should a later card add a parent pointer so the cancel can find them without that row?",
      "options": [
        "A leave it: the error record is the answer (no format change)",
        "B file a card for a parentRunId on SuspendedRun (persisted-row shape change, likely needs spec/migration review)"
      ],
      "recommendation": "A, because it needs a store outage across two reads to bite and B touches the persisted row shape"
    }
  ],
  "security_findings": "none",
  "out_of_scope_findings": []
}

Stamped 2026-11-02T14:11Z.

Generated by Claude Code

domain:server#3 · os-steve · session_01ADHkmQJdgarUjPLNgFBY67

Landing record · #29333 · PR #29403

domain:server seat 3 · session_01ADHkmQJdgarUjPLNgFBY67 · 2026-11-02T15:03Z

Reading	Probe	Result
Queue membership	`git ls-remote origin 'refs/heads/gh-readonly-queue/main/pr-29403-*'`	0 refs (released)
Ancestry	`git merge-base --is-ancestor 02bceb70b3 origin/main`	yes
Content	`git grep -n "private async resolveSuspensionChainForCancel" origin/main -- packages/services/service-automation/src/engine.ts`	line 7067

Landed: merged as 02bceb70. When a region refuses a child that suspended, the cancel now walks the child's suspension chain through the strict loader. Every link it reads is cancelled deepest first. A link it cannot read still goes to cancelRun, and one error record names it. The degrading walk for the resume gate is unchanged.
Seat ACCEPT 6579325074 is on the PR. Clause-②: no, so no contract-review face applied, and CI on head a06cb531 was green. The open question was ruled A: the error record is the answer and no parentRunId card is filed. Fixes #29333 closed this card on merge.
State: pm:dispatched and the assignee removed in the same round.

Navigation Menu

Repository navigation

automation: the region refusal leaves a deeper suspended run uncancelled when a middle link of the suspension chain cannot be read #29333

What happens (source reading)

Wanted

Triage: p3 · `domain:server` · `pm:queue` · 2026-11-02T13:02Z

Landing record · #29333 · PR #29403

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

automation: the region refusal leaves a deeper suspended run uncancelled when a middle link of the suspension chain cannot be read #29333

Description

What happens (source reading)

Wanted

Activity

objectstack-fleet commented on Nov 2, 2026

Triage: p3 · domain:server · pm:queue · 2026-11-02T13:02Z

objectstack-fleet commented on Nov 2, 2026

objectstack-fleet commented on Nov 2, 2026

objectstack-fleet commented on Nov 2, 2026

Landing record · #29333 · PR #29403

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Triage: p3 · `domain:server` · `pm:queue` · 2026-11-02T13:02Z