Reproducibility

Environment

Python 3.10, macOS. Exact package versions: Install.

Known limitations

• — SPARC galaxy rotation-curve data is not present; that test doesn’t run.
• — GRB per-photon arrival-timing data is not present; the multi-messenger pipeline check uses injected synthetic timing on top of real energy/redshift data.
• — The ringdown check validates the analysis pipeline against synthetic data, not a real detector signal.
• — The effective-spin shadow-asymmetry claim requires 2D ray-tracing not implemented in this lab (marked NOT_TESTABLE, not approximated).
• — Group B’s first run reads ~28 GB of pre-existing simulation snapshots — expect several minutes before caching kicks in.

Determinism

Simulations that involve random seeds (e.g. Born-rule measure-concentration Monte Carlo) are not currently pinned to a fixed seed in the published run — re-running may produce slightly different sample-path detail while preserving the same pass/fail conclusion at the stated tolerance.