[aw-failures] Conclusion job's setup_globals.cjs MODULE_NOT_FOUND cascade has regressed (repeat of #28149)

[github-actions]

Problem

Give the conclusion job's checkout a retry, or bundle its helper scripts — this exact cascade was already root-caused and closed as fixed in April, and it just came back. §30483968883 (PR Sous Chef, 19:20 UTC): the agent, safe_outputs, evals, and every other job in the run succeeded — only the conclusion job failed, at its "Checkout actions folder" step, then cascaded into "Handle agent failure":

##[error]fatal: unable to access 'https://github.com/github/gh-aw/': Could not resolve host: github.com
##[error]fatal: could not fetch e3af0082a628e90622317952eb61abf961f575e0 from promisor remote
##[error]The process '/usr/bin/git' failed with exit code 128
...
Error: Cannot find module '/home/runner/work/_temp/gh-aw/actions/generate_usage_activity_summary.cjs'
...
Error: Cannot find module '/home/runner/work/_temp/gh-aw/actions/setup_globals.cjs'
##[error]Unhandled error: Error: Cannot find module '/home/runner/work/_temp/gh-aw/actions/setup_globals.cjs'

This is the identical two-stage cascade described and closed in #28149 (2026-04-23): a transient DNS/network failure during the conclusion job's fresh git checkout for AWF action helpers turns into a hard MODULE_NOT_FOUND, because handle_agent_failure.cjs (and here, also generate_usage_activity_summary.cjs) import setup_globals.cjs from that same checkout with no fallback. #28149's own remediation options (retry-with-backoff on the checkout, or pre-bundling the helper scripts into the harness) were apparently never applied to this checkout step, or the regression reintroduced an unguarded checkout path.

Affected Workflows and Runs

• .github/workflows/pr-sous-chef.lock.yml — conclusion job, "Checkout actions folder" / "Handle agent failure" steps
• §30483968883 (2026-07-29 19:20 UTC) — DNS resolution failure (Could not resolve host: github.com) during checkout, cascading to 2 separate MODULE_NOT_FOUND errors
• Prior identical pattern (closed as fixed): #28149, 8 runs across 4 workflows on 2026-04-23

Note: the agent and safe_outputs jobs of this same run succeeded fully (see closed #48899 — the previously-tracked AWI_MAINTENANCE_TOKEN 401 issue is unrelated and confirmed fixed by this same run). Only the conclusion/reporting path is affected.

Probable Root Cause

Same as #28149: the conclusion job performs a live git checkout to pull in helper scripts (setup_globals.cjs, generate_usage_activity_summary.cjs, handle_agent_failure.cjs) instead of using pre-bundled harness artifacts. A transient GitHub network blip (DNS/HTTP 5xx) during that checkout is fatal with no retry, and the missing checkout cascades into unrelated-looking MODULE_NOT_FOUND errors that obscure the real root cause for anyone reading only the tail of the log.

Proposed Remediation

Reuse the remediation already scoped in #28149 (apparently not fully/durably applied):

1. Preferred: Add retry-with-backoff (e.g. 3 attempts, 5s/15s/30s) around the conclusion job's "Checkout actions folder" step.
2. Alternative: Pre-bundle setup_globals.cjs and other conclusion-job helper scripts into the AWF harness artifacts so the conclusion job never depends on a live checkout.
3. Regardless of which fix path is taken, add a regression test/check so this doesn't silently regress a third time — e.g. an integration test that simulates checkout failure and asserts the conclusion job still reports gracefully instead of throwing MODULE_NOT_FOUND.

Success Criteria

• A transient DNS/HTTP 5xx failure during the conclusion job's checkout no longer produces an unhandled MODULE_NOT_FOUND crash.
• The conclusion job either succeeds after retrying the checkout, or fails with a clear, direct error message about the checkout failure (not a cascaded missing-module error).
• No recurrence of this exact signature in the following 6h window.

Parent: #48897
Related: #28149 (closed, same root cause), #48899 (closed, unrelated credential issue in the same run)
Related to #48897

Generated by 🔍 [aw] Failure Investigator (6h) · age00 · 161.9 AIC · ⌖ 20.7 AIC · ⊞ 5.2K · ◷

• expires on Aug 5, 2026, 5:25 PM UTC-08:00

[github-actions]

🍪 Issue Monster selected this for Copilot

I have identified this issue as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · gpt54 · 3.9 AIC · ⌖ 5.69 AIC · ⊞ 6.4K · ◷