fix(cli): forward function requests without cloning (CLI-2415)

[7ttp]

TL;DR

Prevents early Edge Function responses from stalling locally.

Cloning tees the request body, leaving an unused stream branch that can block early responses,
which we've now fixed by forwarding the body without cloning and covering the failure with a runtime regression test.

ref:

• closes CLI request cloning can stall an early Edge Function response #6564

[github-actions]

🤖 AI Review

The request-forwarding fix is correct, and no runtime correctness or security defect was found. Of Claude’s four findings, two quality concerns are confirmed, one test-reliability concern remains uncertain without reproducing the pre-fix behavior, and the naming concern is refuted. Codex reported no findings.

Findings

Severity	Location	Category	Sources	Claim
🟡 MINOR	apps/cli/src/shared/functions/serve-main-offline.e2e.test.ts:500	test-reliability	claude	The regression test may not reliably exercise the pre-fix body-stream stall, and timeout failures lack container diagnostics.
🟡 MINOR	packages/stack/src/functions/serve.main.ts:198	test-coverage	claude	The stack-owned request-forwarding fix has no regression test covering an early response while its request body remains unread.
⚪ NIT	apps/cli/src/shared/functions/serve.main.ts:291	maintainability	claude	Neither request-forwarding implementation documents the non-obvious requirement not to clone the request, allowing the body-tee stall to be reintroduced.

Refuted findings (kept for transparency, not posted as review comments)

• apps/cli/src/shared/functions/serve.main.ts:288 (naming): The clonedURL identifier is stale because the function no longer clones anything.
  Refuted: The identifier still accurately describes the distinct URL object created from the request URL at line 288; only the Request clone was removed.

Stats

Claude findings: 4 · Codex findings: 0 · Confirmed: 2 · Refuted: 1 · Uncertain: 1

---

Models: claude-opus-5 + gpt-5.6-sol · Trigger: auto · Workflow run

This review runs once per PR. A maintainer can request another with a /ai-review comment.