Five documents describe an answered entry of the plan's question issue as open, and one of them withholds a hand-over the record permits

[iderex]

Every entry of the plan's question issue was answered on 2026-08-24, the issue
closed on 2026-08-27, and eight decision records landed for the eight answers.
Five documents still describe entries of it as open, unanswered or conditional
on an answer that exists. One of them withholds a capability the recorded
answer grants.

gh issue view 46 --repo Flowfin/lab --json state --jq '.state'
CLOSED
git ls-tree --name-only origin/main docs/decisions/ | sed 's#.*/##' | cut -c1-4 | tr '\n' ' '
0000 0001 0002 0003 0004 0005 0006 0007 0008 0009 0010 0011 0012 0013 0014 0015 0016 0017 0018 0019 0020 0021 0022 0023 0024 0025

The sites, derived rather than listed by hand:

git grep -n 'issue #46' -- .
docs/decisions/0001-the-runner-is-written-in-go.md:70:Whether this board publishes downloadable artefacts is entry four of issue #46
docs/decisions/0005-how-a-result-leaves.md:53:three of issue #46 and is not decided here. This record fixes what the section
docs/decisions/0012-the-supported-platforms.md:78:of issue #46 and is open. If the answer there is source only, the three
docs/promotion.md:40:is the answer both of them come from. Entry three of issue #46 asks who may
docs/quality-parity.md:654:question was an entry on issue #46, and that entry is answered:
docs/supply-chain.md:125:outside at all is an open question on issue #46 rather than something to score
docs/supply-chain.md:208:issue #46 asks whether this board publishes downloadable artefacts at all. If the
docs/supply-chain.md:274:signed are both entry four on issue #46.

The one that is not a stale reference

docs/promotion.md:40 tells a reader that a hand-over of somebody else's work
cannot be completed:

Entry three of issue #46 asks who may place a contributor's work under another
board's terms and carries no answer, so a hand-over of somebody else's work
still cannot be completed, and writing anything else into this line would be
inventing permission nobody gave.

That entry is answered and its record is on the default branch:
docs/decisions/0020-consent-to-promote-under-other-terms.md decides that a
result may be promoted under different terms with the contributor's explicit
consent, given at promotion time. The record says of itself that the answer
reaches the checklist and that the section gains a fifth item. It has not:

grep -c -i 'consent' docs/promotion.md
0

So the checklist names four things where the record says five, and the sentence
above refuses a hand-over the decision permits. Of the whole set this is the
only site where the drift costs something other than a reader's confidence.

The three in the supply-chain triage

docs/supply-chain.md:125 calls whether this board takes contributions from
outside an open question on that issue; entry eight answered it - anybody - and
docs/decisions/0024-who-may-run-an-experiment-here.md records it.

docs/supply-chain.md:208 and :274 rest the Packaging and Signed-Releases
rows on entry four, including a branch for the answer being source only.
docs/decisions/0021-what-this-board-publishes.md decides that this board
publishes downloadable artefacts and that they are signed, so that branch cannot
be taken and the two rows have a condition rather than a question.

The one in a decision record, which is a question rather than an edit

docs/decisions/0012-the-supported-platforms.md:78 states the entry is open:

It does not decide whether this board publishes downloadable release artefacts
at all, which is entry four of issue #46 and is open.

CONTRIBUTING.md says a record is replaced by a later record rather than edited,
so what repairs a landed record that carries a claim which has stopped
reproducing is not settled here. The answer was not source only, so the
conditional this record makes itself superseded by never fires and the rest of
it stands; what is wrong is three words.

Two sites that are NOT wrong

docs/decisions/0001-the-runner-is-written-in-go.md:70 and
docs/decisions/0005-how-a-result-leaves.md:53 name entries of that issue and
say the question is not decided in the record carrying the sentence. Both
statements are about their own record's scope and both are still true, so
neither is in this issue's scope. Record 0020 already names 0005 and says
what moved.

How it was found

By running the commands a neighbouring document pastes before quoting a row back,
and then asking the same question of every document that names the same issue. A
claim about another artefact reads the same whether or not the artefact still
says it, which is why nothing about any of these paragraphs looks wrong.

Done when

docs/promotion.md carries the fifth checklist item record 0020 says the
section gains, and no longer says a hand-over of somebody else's work cannot be
completed; the three rows in docs/supply-chain.md read against the answers
rather than against an open question; the sentence in
docs/decisions/0012-the-supported-platforms.md is repaired by whatever route
this board decides a landed record is repaired by, with that route written down;
and git grep -n 'issue #46' returns no site asserting an entry is open.

[iderex]

The two documents are repaired and the record sites are not

#217 lands the prose half, merged as 55ef1616e1da8889c005fb7f185fae613557e839
with all twenty-five checks green. docs/promotion.md carries the consent item
record 0020 says the section gains, together with the residual that nothing
here refuses a section naming no consent; the three rows in
docs/supply-chain.md read against records 0021 and 0024 rather than against
an open question. No re-score was run and neither row claims one.

git grep -n 'issue #46' -- docs/promotion.md docs/supply-chain.md
exit=1

A second record carries the same class, and it was found in the same walk

The issue body names docs/decisions/0012-the-supported-platforms.md as the one
site whose repair is a rule question. There are two.
docs/decisions/0018-the-licence-of-this-board.md carries three claims that have
stopped reproducing, in one paragraph.

It says the file on the default branch says something else, and pastes:

gh api repos/Flowfin/lab --jq '.license.spdx_id'
GPL-3.0

It says the answer and the file disagree until #47 lands the file. That issue is
closed and the file is on the default branch:

gh issue view 47 --repo Flowfin/lab --json state --jq '.state'
CLOSED
head -1 LICENSE
                    GNU GENERAL PUBLIC LICENSE

And it says what this repository declares is empty, pasting the constant:

git grep -n 'const DeclaredLicence' origin/main \
  -- internal/invariants/invariants.go
origin/main:internal/invariants/invariants.go:92:const DeclaredLicence = "GNU GENERAL PUBLIC LICENSE"

The paragraph closes by telling a reader arriving between the record and that
change to trust neither, which is exactly right for the day it was written and
is what makes the drift readable rather than harmless: a reader who follows it
today is told to distrust a file that now agrees with the record.

Why this makes the rule question harder to leave

Two landed records now carry claims about other artefacts that no longer
reproduce, and record 0000 gives a landed record exactly one legal edit, the
line naming what superseded it. Neither of these is a change to what was decided

• 0018 still decides GPL-3.0 and 0012 still fixes the six platforms - so
  superseding either would tell a reader the ground moved when it did not.

What is being asked for is the route by which a landed record's evidence is
brought up to date, or a written decision that it never is and that a reader is
expected to re-run the pastes. Either answer is workable and neither is mine to
take, so the label goes on and this waits.

Done when, as it now stands

The route above is decided and written down, and both sites are left in whatever
state that route says. The prose half is landed and is not part of what is left.

[iderex]

What this waits on

the route by which a landed decision record's stale evidence is brought up to date, or a written decision that it never is: records 0012 and 0018 both carry pastes that no longer reproduce, and record 0000 gives a landed record only one legal edit

[iderex]

Deciding the route, and it is the one this fleet already lives by elsewhere: a landed record is never brought up to date, because its pastes are measurements of their date, not claims about today - editing them would forge a reading nobody took. Record 0000's single legal edit stays exactly as narrow as it is. When a paste stops reproducing, one of two things is true. Either the decision still stands on today's evidence, and then nothing happens - a stale paste under a standing decision is history, not error. Or the world moved far enough that the decision would fall differently today, and that is never an edit: it is a new record superseding the old one, carrying the fresh measurement and naming what it replaces, with the old record left word-stable as the reason things were once otherwise. Records 0012 and 0018 get read against that test by whoever picks this up: standing decision, no action; fallen premise, successor record. Unblocking.

[iderex]

Closed with one clause of the original done-condition deliberately unmet

The route is written down. docs/decisions/0026-evidence-in-a-landed-record-that-stopped-reproducing.md is on the default branch at 9208ceb599a294328bde3d8b660c65a5fd3c5fb5, from #219 with all twenty-five checks green: a landed record's evidence is never brought up to date, because a paste is a measurement of its date rather than a claim about today. Where the decision still stands on today's evidence, nothing happens. Where the premise fell, the answer is a successor record, never an edit.

Records 0012 and 0018 were read against that test and both stand, so neither is edited and neither gains a successor. 0012 makes itself superseded only if entry four of #46 answered source only, and record 0021 decides that this board publishes signed downloadable artefacts, so its conditional cannot fire. 0018 decides GPL-3.0, and the three readings it said disagreed now agree with it.

The clause that is not met is the last one in the body: that git grep -n 'issue #46' return no site asserting an entry is open. One still does, and under the route above it stays:

git grep -n 'issue #46' -- . | cut -d: -f1,2
docs/decisions/0001-the-runner-is-written-in-go.md:70
docs/decisions/0005-how-a-result-leaves.md:53
docs/decisions/0012-the-supported-platforms.md:78
docs/quality-parity.md:654

The first two are each about their own record's scope and the body puts them outside this issue; the fourth says the entry is answered. The third is 0012, and repairing it is what the route refuses. That clause was written before the route existed and the route is what overturns it, which is why this is written here rather than left for a reader to reconstruct from a closed issue.

I am naming that rather than closing quietly, because a done-condition dropped in silence is indistinguishable from one nobody read.