[SDK Sentinel] Stabilize unfinished-handler cancellation test

[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 · first known failure

Stabilize the unfinished-handler cancellation test after the [first captured macOS failure](https://github.com/temporalio/sdk-python/actions/runs/34256407686/job/102163229591) and a [corroborating failure](https://github.com/temporalio/sdk-python/actions/runs/34262268385/job/102195905654). No equivalent last-known-good run was included in the snapshot.

Root cause

The setup requested cancellation before starting the update. Server scheduling could advance cancellation first, leaving the update RPC to observe a failed workflow task instead of joining the workflow task that the test intends to exercise.

Fix

Queue the update or signal first, wait until it is accepted, then request cancellation before starting the worker. This preserves the intended shared-workflow-task scenario while removing nondeterministic request ordering.

Validation

The pre-change focused run did not reproduce the failure. After the change, all 48 focused parameters passed, as did 40 repeated cancellation/update cases. The canonical lint, type, bridge-lint, and editable-build validator passed. macOS Python 3.14 CI remains required because local validation ran on Linux.

Validation status: validated-without-reproduction

• Flake confidence: high
• Fix confidence: high
• Value: medium
• Patch scope: standard — 1 changed file; 984 B
• Local reproduction: not-reproduced — One pre-change focused execution of all 48 parameters passed; the CI race did not occur locally.
• Regression coverage: pass-after-only — Existing parametrized coverage passed after the change, including 40 repeated cancellation/update cases; no fail-before reproduction was obtained.
• Unchanged baseline (Poe lint, bridge lint, and editable build) — passed
• Independent candidate (Poe lint, bridge lint, and editable build) — passed

Investigator-run checks

• poe test -s -k test_unfinished_handler_on_workflow_termination — passed; 2 attempt(s). All 48 parameters passed before and after the change; the pre-change run did not naturally reproduce the flake.
• poe test -s -k 'test_unfinished_handler_on_workflow_termination and cancellation and update' --flake-finder --flake-runs=5 — passed; 1 attempt(s). All 40 post-change cancellation/update stress cases passed.
• .ci-flake-runtime/input/validate.sh — passed; 1 attempt(s). Ruff, Pyright, mypy, basedpyright, pydocstyle, bridge Clippy, and editable bridge build passed.

Required target CI

• Continuous Integration / build-lint-test (3.14, macos-arm)

Residual risks

• The stress check ran on Linux rather than the observed macOS runner.
• Extreme server load could still expose a different workflow-task readiness failure.

Automation provenance

• Investigator workflow
• Base SHA: ee2a4da7b075b96e574d4388d27f61fb9d205ab8

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: 1
• First occurrence: 2026-09-08 17:36 UTC
• Latest occurrence: 2026-09-08 17:36 UTC
• Recent occurrences:
  • 2026-09-08 17:36 UTC — failed CI job