Results

The write-ups above cite numbers; these are the files those numbers come from, kept in the repository rather than copied here so they can't drift.

Compatibility

compat/README.md — how the sweep works, and every environment variable it reads.
compat/results/v0.21.0.md and one file per earlier release — the sweep's install and lock-resolve comparison for that tag, pinned corpus and random sample both.
compat/hunted.md — public projects swept once outside the pinned corpus, so a repeat hunt starts from what's already covered.
compat/results/lock-age.md — how often a lock 1, 2 or 4 years old still installs today (candidate B).
compat/results/platform-drift.md — how often a locked package's own PHP floor drifts from the root's claimed support (candidate D).
compat/results/g3-plugins.md — what the most-downloaded Composer plugins actually do at install time (candidate 3.4).

bench/results/README.md — the headline cold/warm/no-op/update table against Composer and riff.
bench/results/corpus.md — the same scenarios across ten corpus projects, including vivacity.
bench/results/profile.md — where install and update spend their time, flamegraphs included.
bench/results/lockmerge.md — the merge-conflict replay behind chapter 1.
bench/results/workspaces.md — the aggregate-root measurements behind chapter 3.
bench/results/storeload.md — boot cost and disk use of a linked vendor/ against the store (chapter 2).
bench/results/g3-toolchain.md — how often the corpus needs a PHP version or extension the host lacks (candidate 3.1).
bench/results/g3-isolation.md — what popular plugins bundle their own Composer dependencies (candidate 3.2).
bench/results/g3-isolation-sites.md — how often a plugin's bundled library clashes with the site's own copy, on five Composer-managed WordPress sites (candidate 3.2 follow-up).
bench/results/g3-rules.md — a feature census behind simplifying resolution rules (candidate 3.5).