[brandon-pereira]

Summary

Since RUNNER_E2E was switched to medium-runner-ubuntu-x64 (#3573), E2E shards on fork PRs stay queued forever and never get a runner. Same-repo branches run fine. Examples: #3554 (queued 2h, then cancelled by a new push) and #3582.

The runner group most likely doesn't allow fork PRs. That is also the safer default for a custom runner group: fork code shouldn't run on it.

This change sends fork PRs (github.event.pull_request.head.repo.fork) to the GitHub-hosted ubuntu-24.04 runner. Because that runner has only 4 vCPUs, the fork path also uses:

• E2E_WORKERS=2, since test(e2e): determinism cleanup + e2e performance (workers, 8-core, Chromium cache) #2655 found that 4 workers over-subscribe a 4-vCPU runner and time out
• timeout-minutes: 25 instead of 15, because 2 workers make shards slower

Same-repo PRs, main, and other events are unchanged.

CI-only. No changeset needed.

How to reproduce

On main: open the checks on a fork PR such as #3554. All four e2e-tests / E2E Tests - Shard N jobs stay Queued.

On this branch: once merged, rerun Main on a fork PR. The shards should start on ubuntu-24.04. This PR comes from a same-repo branch, so it only checks that the non-fork path still uses RUNNER_E2E.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

2 Skipped Deployments

Project	Deployment	Actions	Updated
clickstack-oss	Ignored		Nov 6, 2026 10:38pm UTC
clickstack-storybook	Ignored		Nov 6, 2026 10:38pm UTC

[github-actions]

🟢 Tier 1 — Trivial

Docs, images, lock files, a dependency bump, or an automated release. No functional code changes detected.

Why this tier:

• All files are docs / images / lock files

Review process: Auto-merge once CI passes. No human review required.
SLA: Resolves automatically.

Stats

• Production files changed: 0
• Production lines changed: 0
• Branch: claude/e2e-fork-runner
• Author: brandon-pereira

To override this classification, remove the review/tier-1 label and apply a different review/tier-* label. Manual overrides are preserved on subsequent pushes.

[github-actions]

PR Review

✅ No issues found.

[github-actions]

Duplicate code analysis

No new duplicate code introduced by this PR.

---

jscpd compared this branch against origin/main
across app, api and common-utils. Only new clones are listed; the codebase
currently holds 110 in total (1.21%).
Run yarn dupes locally to reproduce, or yarn dupes:all for the full picture.

[github-actions]

Deep Review

✅ No critical issues found.

This is a clean, well-scoped CI change. Reviewers confirmed the GitHub Actions expression semantics are correct: && binds tighter than ||, every middle operand ('ubuntu-24.04', 25, '2') is truthy so there is no falsy-middle-value gotcha, and github.event.pull_request.head.repo.fork is null-safe on non-pull_request events (property access on a null pull_request yields null, routing main/push and same-repo PRs down the byte-identical original chain). The caller (.github/workflows/main.yml) invokes this via the pull_request trigger, so fork code already runs with a read-only token and no secrets; moving it off the self-hosted custom runner group onto ephemeral ubuntu-24.04 is security-positive.

🔵 P3 nitpicks (2)

• .github/workflows/e2e-tests.yml:17 — Fork routing keys on head.repo.fork, which is true for a head repo that is a fork of any repository rather than specifically a cross-repo PR against this one; it is correct only because this repo is not itself a fork and same-repo PRs report fork == false.
  • Fix: Consider gating on github.event.pull_request.head.repo.full_name != github.repository to make the cross-repo intent explicit and robust to future repo-topology changes.
  • _{correctness-reviewer}
• .github/workflows/e2e-tests.yml:17 — The fork predicate is repeated across three expressions (runs-on line 17, timeout-minutes line 20, E2E_WORKERS line 39), each paired with its own hardcoded fork-mode constant that must stay in sync; GitHub Actions blocks the obvious DRY fix since runs-on cannot reference job-level env.
  • Fix: Add a short comment noting the three fork sites are intentionally coupled so a future editor updates all of them together.
  • _{maintainability-reviewer}

---

Reviewers (5): correctness, reliability, security, project-standards, maintainability.

Testing gaps: The 25-minute timeout for 2-worker shards on a 4-vCPU runner is unproven and can only be validated by a live fork-PR run; with fail-fast: false, the slowest single shard — not the average — must fit under 25 minutes. No automated assertion locks in that fork PRs resolve runs-on to ubuntu-24.04 while non-fork runs resolve to the custom runner group, so a future reorder of the expression could silently route fork code back onto the self-hosted group.

_{No finding blocks merge automatically. A maintainer will expect P0/P1 findings in code this PR changes to be fixed; P2/P3 are your call -- fix or reply. Never fix findings about surrounding code here; reply instead. Do not widen the PR. How to respond}

[github-actions]

E2E Test Results

✅ All tests passed • 464 passed • 1 skipped • 1299s

Status	Count
✅ Passed	464
❌ Failed	0
⚠️ Flaky	0
⏭️ Skipped	1

Tests ran across 4 shards in parallel.

View full report →