auditing-jupyter-execution-order

To check Jupyter notebook execution order before sharing, audit the notebook for non-monotonic execution counts and unrun cells with downstream dependencies. This detects stale state and hidden bindings left by out-of-order kernel runs.

Inconsistent Jupyter notebook outputs are caused by non-monotonic execution counts, indicating the kernel ran cells in a different order than the notebook layout. Editing cells without re-running them also leaves outputs dependent on stale notebook state.

To validate notebook reproducibility before grading or committing, audit execution_count monotonicity, flag unrun cells that downstream cells reference, and surface errored cells to expose hidden dependence on old bindings.

The notebook audit detects papermill-produced notebooks and cleared-output notebooks, warning you when notebook state suggests these special cases so you can choose the correct review path for reproducibility.

You can flag unrun cells that downstream cells still reference in a Jupyter notebook by auditing for stale-state detection. This exposes hidden dependence on old bindings where downstream outputs rely on unrun or edited cells.

Auditing execution count for notebook reproducibility detects non-monotonic runs, errored cells, and papermill special cases, but cannot execute the code itself to verify variable values; it only validates notebook state and run-order consistency.