friction: stabilize ci_uv_sync_diag timeout under shared-host load

[ll7]

Archetype Metadata

archetype: test
evidence_tier: smoke
linked_policy: []

Goal / Problem

Make the CI ci_uv_sync_diag validation test reliable under the repository's normal shared-host
load. During the #8237 final readiness run, the diagnostic subprocess exceeded its 30-second test
timeout even though the same test module passed immediately when isolated.

Skipped or failing test evidence

• Test path: tests/dev/test_ci_uv_sync_diag.py::test_ci_uv_sync_diag_reports_runner_and_uv_state
• Failing command: PR_READY_MODE=final BASE_REF=origin/main scripts/dev/pr_ready_check.sh
• Observed result on 2026-09-01: the core lane reported one failure after the diagnostic subprocess
  timed out at 30 seconds; 5,653 other tests passed and 3 were skipped.
• Isolated result: uv run pytest tests/dev/test_ci_uv_sync_diag.py -q passed all 7 tests in 2.46s.

Inputs

• tests/dev/test_ci_uv_sync_diag.py
• scripts/dev/ci_uv_sync_diag.sh
• The friction: reconcile assurance-case evidence registry baseline drift #8237 readiness transcript and the isolated 7-test diagnostic run.

Intended behavior

The diagnostic test should complete reliably within the repository's supported validation
environment, or its timeout/resource policy should fail with a clear classification rather than
creating an ambiguous full-gate failure.

Scope

In scope:

• Identify whether the timeout is caused by subprocess startup, uv/filesystem contention, or an
  undersized test timeout.
• Add deterministic regression coverage or an appropriately bounded timeout/resource guard.

Non-goals:

• Changing the friction: reconcile assurance-case evidence registry baseline drift #8237 assurance-ratchet hotfix.
• Weakening the diagnostic contract or making the readiness gate advisory.
• Killing or coordinating unrelated jobs on shared hosts as a test workaround.

Acceptance criteria

• The full readiness lane no longer fails spuriously on the diagnostic timeout under supported
  shared-host load.
• The isolated diagnostic test and its relevant CI script contract tests pass.
• Any timeout increase or retry is justified by measured startup/resource behavior and remains
  bounded.
• The result distinguishes a real uv/runner failure from host-contention noise.

Targeted validation

uv run pytest tests/dev/test_ci_uv_sync_diag.py tests/test_ci_script_contract.py -q
BASE_REF=origin/main scripts/dev/pr_ready_check.sh

Artifact policy

Evidence category: disposable validation output; no benchmark or release artifact is produced.

Keep output/ logs and coverage files untracked and worktree-local. Promote only a small
reproducible fixture or report if the diagnosis establishes a durable contract.

Blocked / ready state

blocked by flaky environment

Estimate metadata

Effort (h): 2
Complexity: low to medium
Risk: low; test-infrastructure-only, but timeout changes can mask real failures
Uncertainty: medium; the isolated pass suggests shared-host contention, but the exact bottleneck is
not yet measured.

Validation / Testing

• Reproduce or characterize the timeout with bounded shared-host load.
• Run the targeted diagnostic and script-contract tests.
• Run the canonical readiness gate and record whether the result is clean.

[ll7]

Triage audit — needs-triage cleared

Ran the canonical template audit (scripts/tools/issue_template_audit.py) against the live body:

• all required sections present; missing_sections: [];
• ## Archetype Metadata block present with valid archetype: test, evidence_tier: smoke; no missing keys or invalid values.

No body repair needed. Removing needs-triage; readiness gating follows.

[ll7]

implementation-claim: opencode-goal-autopilot-20260904 @ 88b7188 until 2026-09-04T02:00:00Z