[sdk-sentinel-bot]

Caution

This PR contains untrusted AI-generated code. Do not approve or run CI until a maintainer has reviewed the diff. SDK Sentinel verified that GitHub Actions remained approval-gated with zero executable jobs when this PR was opened.

Summary

• CI evidence: failed CI job · workflow run · last known pass · first known failure

Stabilize the OpenAI Agents tracing integration test after a main Python 3.14 failure. A [preceding equivalent suite](https://github.com/temporalio/sdk-python/actions/runs/34913775343/job/104206831444) passed the test, while the [first captured failure](https://github.com/temporalio/sdk-python/actions/runs/35031772047/job/104591666764) observed two span-start callbacks where an adjacent start/end pair was expected.

Root cause

Workflow execution may begin before the client start-workflow RPC span finishes. The test paired callbacks by list position, so valid asynchronous interleaving made it compare two start events.

Fix

Select the startWorkflow callbacks by semantic name and span identity, then retain the existing hierarchy assertions. Paired callbacks now also must share a span ID.

Validation

The unchanged exact test passed 6/6 attempts, so natural reproduction was not obtained. The candidate passed 11/11 focused attempts. The trusted baseline and fixed repository validator both passed. The originating Ubuntu matrix job remains required.

Validation status: validated-without-reproduction

• Flake confidence: high
• Fix confidence: high
• Value: medium
• Patch scope: standard — 1 changed file; 7.7 KiB
• Local reproduction: not-reproduced — Six unchanged-checkout focused attempts passed; the CI log supplies the observed callback interleaving.
• Regression coverage: pass-after-only — The existing exact test passed 11 candidate attempts; no new test was needed because the change strengthens its span-identity assertions.
• Unchanged baseline (Poe lint, bridge lint, and editable build) — passed
• Independent candidate (Poe lint, bridge lint, and editable build) — passed

Investigator-run checks

• uv run pytest tests/contrib/openai_agents/test_openai_tracing.py::test_tracing -q — passed; 11 attempt(s). Ten post-change stress attempts and one post-format attempt passed.
• .ci-flake-runtime/input/validate.sh — passed; 1 attempt(s). Ruff, type checks, docstyle, bridge lint, and editable bridge build passed.

Required target CI

• Continuous Integration / build-lint-test (3.14, ubuntu-latest)

Residual risks

• The rare callback interleaving did not occur locally.
• A distinct Linux ARM parent-cardinality assertion in the same test remains outside this signature and candidate.

Automation provenance

• Investigator workflow
• Base SHA: f74399de1c267c22c69f7713c66ea4a332465ad8

If this finding should not be fixed, apply the sdk-sentinel:false-positive label and close the PR. Sentinel will suppress the finding until its affected source changes.

SDK Sentinel recurrence history

• Unique failed CI run attempts: 2
• First occurrence: 2026-09-15 17:17 UTC
• Latest occurrence: 2026-09-15 22:55 UTC
• Recent occurrences:
  • 2026-09-15 22:55 UTC — failed CI job
  • 2026-09-15 17:17 UTC — failed CI job

[brianstrauch]

We're in the process of migrating this package to https://github.com/temporalio/ai-integrations. I'll accept these changes over there once the migration is complete.