The footer census is a command, and the chapter that guessed it says so

The seed's own measurement of its sovereignty footer was taken with the wrong instrument, and this release replaces the number with the command that measures it.

What was wrong. §8 of SEED.md told a reseeding line that the footer appears on four of forty releases, and built on that a sentence about alarm fatigue. The check the paragraph offered as its own proof, git tag -l --format='%(contents)', reads tag messages; the convention is about release notes, which are not in this repository at all. It returned four, and kept returning four, because no annotated tag had been cut since. A wrong instrument is stable; re-running it is not a second measurement. The sentence ran from v1.44.0 through v1.61.0, nineteen releases.

What ships.

• SEED.md §8 carries no live count. It carries the census command, which reads release notes through the API, with its edge cases stated: pre-convention releases, tag-only versions, drafts, exact-match comparison. The false rate and the alarm-fatigue sentence are struck in place, dated.
• HISTORY.md holds the figures as a dated snapshot, taken 2026-09-02 before v1.61.0 was cut: across all sixty-six releases, 48 last, 3 first, 15 absent. Release notes are editable after publication and the forge keeps no history of a body, so any census measures the current record.
• CORRECTIONS.md gains C-11, admitted on harm, addressed to anyone who read §8 at any release from v1.44.0 through v1.61.0: we asserted as measured fact something we had not measured, in the chapter that builds a line's daily reseed routine.
• CONTRIBUTING.md and ADOPTING.md carry dated strikes where they had inherited the conclusion.

Every cold read of this branch after the first blocked inside the previous revision's repair, and what finally converged was deleting claims rather than rewording them. Closes #54, #62 and #64.

Take what fits, item by item; nothing here can weaken a floor.