Historical records

This section preserves migration records, backend postmortems, hardware evidence, architecture decisions, and kernel histories. For current usage, start with the Python workflow. For current development, use the Developer guide.

Maintainer pages are intentionally detailed. They preserve unsuccessful experiments and exact historical numbers so future work does not repeat a regression or detach a benchmark from its source revision.

Dated investigations and their original open questions live under Historical experiments. They are retained evidence, not the current roadmap.

Detailed records are grouped in the expandable navigation beneath this page.