Dated write-ups of what happened when this project's checkers were pointed at real data.
A finding is not a changelog entry and not a roadmap item. It is a record of a result the method produced against inputs that were not built to make it look good, written so a reader can decide for themselves whether the method earned its conclusion. Negative and partial results belong here on exactly the same terms as positive ones; a robustness check that only ever gets published when it passes is not a robustness check.
Each entry states its own caveats in the body rather than in a footnote, and says plainly which numbers are reproducible from committed inputs and which are not.