Read the catalogue address in the runbook as it answers

[iderex]

What was wrong

docs/domain-cutover.md is the sequence somebody follows while moving the name,
and step 9 is the step that reads the paths nothing links to. It told that reader
that a 404 on the catalogue address is what the address answers today and is
not evidence the move went badly.

The address answers now:

for p in /manifest.json /design-tokens.json; do
  printf "%-40s " "https://flowfin.dev$p"
  curl -sS -o /dev/null -w "%{http_code}\n" "https://flowfin.dev$p"
done
https://flowfin.dev/manifest.json        200
https://flowfin.dev/design-tokens.json   200

Run 2026-08-23. So the sentence inverted the reading it exists to help. On the
day of the move a 404 there is the move having dropped a file, which is the
failure #67 refuses a publish over, and the runbook said to read past it. The
tree's own reading of that address is in harness/needs-network/record/2026-08-23T03-23-54Z.md.

What this change does

git diff --name-only origin/main...HEAD
docs/domain-cutover.md

Three statements, and they are not repaired the same way.

Step 9 now says both paths answer 200 today, so a 404 at either is a dropped
file. It keeps what a 404 used to mean, with the date it stopped meaning it,
because a reader who is given only the new sentence cannot tell why the old one
was ever right.

The paths section said /manifest.json does not answer today. It now says what
it answers and that this repository does not author what it carries, which is
the position record 0006 puts that file in and the reason it is carried rather
than built.

The block under ## The state before the change keeps its 2026-08-11 reading
exactly as printed. That section exists so a rollback restores a state rather
than guessing at one, so editing the snapshot would remove the thing it is for.
A paragraph beside it names the single line that has been overtaken and what the
path answers now.

The means

Prose, in the document that already holds this sequence. What changed is what a
reader is told on the day, and there is nothing here a check could decide: no
reading of this tree says what a public name answers, which is why that class
lives in the harness outside the gate rather than in a leg.

What this does not do

#67 stays open. The publish job that refuses to complete without the
carried file does not exist, and this changes a document rather than building
one.

It edits no decision record. decisions/0006-what-answers-at-the-catalogue-address.md
and decisions/0008-the-url-shape.md both paste the same 404. Under the rule in
decisions/README.md a change to 0006 is not a correction, because 0006 rests on
the address not answering when it says the question is cheap to settle now, and
changing that changes what a reader would decide after reading it. That needs a
new record and is not taken here.

The gate

go run . ci
gate: 7 legs, in order: format, vet, test, build, links, sitemap, invariants
  format: ok, 54 file(s)
  vet: ok
  test: ok, 26 test file(s)
  build: ok, 8 file(s)
  links: every reference that stays inside this site resolves to a file the build wrote
  sitemap: every page the build wrote is listed once, and every entry has a page behind it
  invariants: ok, 38 rule(s) decided, 1 owed and not decided
7 of 7 legs ran. None was skipped.

go run . hygiene origin/main HEAD
hygiene: 1 non-merge commit(s) in origin/main..HEAD, origin internal
  6438a87ab50b: subject carries its reference
1 commit(s) judged, none refused.

Both run 2026-08-23 at 6438a87.

No second reader

This change carries no second reader. The commands above are what stands in place
of one.