The baseline is re-measured against the tree that includes slice G, and this
time exactly one number moved: the **shadow scope** went from **893** measured
files to **894**. The shadow scope is the counted files the repository
deliberately does not gate, and slice G added one of them — a new root test
file, `tests/unit/release/production-stryker-test-files.test.ts`, seven cases
that pin the production path's refusal to hand a missing test file to Stryker.

The generator says out loud that the number moved up, and it is right to: it is
a **count of what is there**, not a ceiling, and nothing is permitted to grow
behind it. Every ceiling held. The lint findings stand at 767, the errors at
257, the over-cap files at 34 and the excess lines at 11430 — all four
unchanged, which is the part worth reading twice: the file that arrived is
**233 raw lines** against the 500-line cap for `tests/`, and it carries no
eslint finding at all, so it enlarges the population without enlarging the
debt.

The enforced scope did not move either. It is still 1383 files, with the lint
ceilings at 1670 findings and 553 errors, the comment rows at 21 narrative
lines and 0 dead references, the silent-warning rows at 44 and 59, and the
prettier and tsc rows at zero. The file-size ceiling still counts 77 enforced
files over the cap and 21825 excess lines, and the hook scope still counts 0 of
61.

This is the third measurement of this baseline in this batch, and it is the
one that belongs to the final tree: the first was taken before slice F was
committed, the second after it, and this one after slice G — which is the only
one whose tree also contains the new test file. The generator's own warning is
the record of why a baseline has to be measured against the tree that exists
rather than the one that was planned: **893 was not wrong, it was stale.**