[anth-volk]

Fixes #935

Same-Repository Draft PR Check

• This PR is a draft.
• This PR is from a branch in PolicyEngine/policyengine-sim-api, not a personal fork.

Summary

Fifth of nine stacked PRs; it builds on #933. It splits the executor's tests/test_stage12_runtime.py (4,709 lines) into test files under the 1,500-line limit for tests, without changing any test or any source file, and removes it from the file-length baseline (count 8 to 7).

A script compared every top-level statement of the old file with the new files: all 168 appear exactly once and unchanged apart from the helper renames, in their original order.

File	Lines	Holds
stage12_runtime_support.py	369	IDs, input and record builders, the coordinate_report wrapper, recording runtimes, failure-summary checks, the dataset-load harness
stage12_fake_stores.py	207	FakeStore, FakeArtifacts and the simulation frames they hold
stage12_child_call_support.py	411	fake child calls and invokers, and the cancellation and log checks
test_stage12_runtime.py	309	how a simulation loads its dataset and region
test_stage12_dataset_provenance.py	335	the dataset a calculation loads, and the provenance it checks
test_stage12_single_worker.py	233	the single worker's successful and refused runs
test_stage12_single_worker_failures.py	337	how the single worker records and reports failures
test_stage12_aggregate_reports.py	454	the aggregate report
test_stage12_coordinator.py	414	starting, aggregating and comparing
test_stage12_coordinator_failures.py	529	a failing child or artifact
test_stage12_coordinator_failure_records.py	403	failure records, dispatch failures, reuse and timeouts
test_stage12_runtime_store_retries.py	135	retrying refused database connections
test_stage12_child_cancellation.py	412	cancelling unfinished children
test_stage12_child_state_errors.py	401	child state errors never replace the recorded failure
test_stage12_wait_window.py	152	the shared wait window

Renames

The helpers the files share are now public, under names no test already uses for a local variable or parameter (a script checked): _spm_calculation_provenance to spm_calculation_provenance, _bundle to bundle_provenance, _simulation to simulation_execution_input, _output_plan to shared_output_plan, _planned_simulation to planned_simulation, _report to report_execution_input, _aggregate_payload to aggregate_payload, _context to invocation_context, _parent to parent_record, _child to child_record, _automatic_parent to automatic_parent_record, _frames to simulation_frames, _connection_limit_error to connection_limit_error, _correlation_id to summary_correlation_id, _failure_cause to summary_failure_cause, _stage12_load to run_until_dataset_loaded, _simulation_id to role_simulation_id, _report_with_segmentation to report_with_segmentation, _coordinate_with to coordinate_with, _assert_cancelled to assert_cancelled, _assert_original_coordinator_failure to assert_original_coordinator_failure, _coordination_log to coordination_log_record, _assert_log_identifies_child to assert_log_identifies_child. Fakes whose unused parameters were named _simulation or _report keep those parameter names.

Seven other test modules imported these helpers from test_stage12_runtime.py (test_stage12_cache_runtime.py, test_stage12_coordination.py, test_stage12_qualification.py, test_stage12_segment_calculation.py, test_stage12_segment_fanout.py, test_stage12_segmentation_foundation.py, test_stage12_simulation_worker.py). They now import them from the support modules under the new names, and test_stage12_coordination.py imports coordinate_report_impl from policyengine_simulation_executor.stage12_runtime directly.

Tests

• 1,895 tests are collected before and after. Every test keeps its name and body apart from the renamed helpers, with one visible difference: pytest names two parametrized cases of test_installed_bundle_accepts_the_dataset_a_region_loads after the helper passed as their parameter, so [us-_bundle] and [state/ut-_bundle] are now [us-bundle_provenance] and [state/ut-bundle_provenance].
• No patch changes: the tests patch the same source modules as before.

Environment variables

None: only test files change.

Testing

• Executor, full suite: 1,873 passed, 22 skipped, as before; the Stage 12 tests (562) were run again on the final layout.
• Ruff 0.15.22 format and lint: clean, including ruff format --check on the executor's tests.
• python scripts/run_quality_guards.py and python -m scripts.rules guard --base-ref origin/file-length-split-stage12-coordination: passed; python -m scripts.rules render leaves no diff.
• Pyright doesn't check test files (each package's include is src), and no source changes.

Only the lowest PR in the stack runs full CI, so this PR's CI complete check fails by design until the PRs below it merge.

🤖 Generated with Claude Code