Is your feature request related to a problem

When multiple dbt Cloud jobs share the same project, a test node can appear in run_results.json from a job that only runs dbt run (models only). dbt includes these tests as compiled-only entries — no SQL is executed — but marks them status: "success". The OM connector ingests this as a real TestCaseStatus.Success, producing phantom test results.

This causes:

• Duplicate results — the same test gets two entries per day (one real from the dbt build job, one phantom from the dbt run job)
• Duplicate alerts — each result triggers a separate change event and notification
• Misleading test history — the timeseries graph shows conflicting statuses (e.g. Success + Aborted for the same test)

How to identify compiled-only entries

Describe the solution you'd like

In add_dbt_test_result() (metadata.py#L1036), skip entries where no SQL was executed:

# After fetching dbt_test_result, before status mapping:
if dbt_test_result.failures is None and not dbt_test_result.message:
    logger.info(
        f"Skipping compiled-only test result for node: {manifest_node.name}"
    )
    return

A real test always populates failures (even 0 on pass) and message. Both being null is a reliable signal that the test was only compiled.

Additional context

• OM version: 1.11.8
• dbt Cloud connector, dbt Core 1.7+
• Related: [Feature] dbt connector: fall back to test_metadata.kwargs.column_name when top-level column_name is null #26606