Parked: ready for BetterVoting when the PR freeze lifts

Status: 🚦 PR FREEZE — Adam was asked not to open new PRs on Equal-Vote/bettervoting until Arend catches up with the existing queue (~60 open, ~43 from the docs/i18n/fixes program indexed in issue #1556).

Everything below is finished or fully specified work, parked here so it survives until publishing is welcome again. Each item carries a flag:

• READY — the change is written and verified; publishing is mechanical.
• READY-AFTER-MERGE — written, but its target only exists once the named PRs land.
• AWAITING-DIRECTION — offered upstream; needs a maintainer to pick an option first.

Do not open any of these as PRs while the freeze is on.

---

1. Re-parent the 31 help pages into the tree — READY-AFTER-MERGE (#1543 + the content PRs)

Every content PR carries parent: BetterVoting Documentation so it could merge independently. Once #1543 and the content PRs land, one follow-up PR applies this map (verified against a full local Jekyll build of all branches merged — 31 pages re-parented, none lost from the nav):

Pattern per page: set parent: and nav_order: as below, and add grand_parent: (For Voters / Setting Up / Counting and Results / Reference pages get grand_parent: BetterVoting Documentation for the two top sections' children, grand_parent: Running an Election for the three subsection's children).

Page (stem in docs/help/)	parent	nav_order
how_to_vote	For Voters	1
after_you_vote	For Voters	2
is_my_vote_secret	For Voters	3
voter_troubleshooting	For Voters	4
how_voters_get_access	For Voters	5
using_another_language	For Voters	6
election_or_poll	Setting Up Your Election	1
choosing_a_voting_method	Setting Up Your Election	2
electing_more_than_one	Setting Up Your Election	3
ballot_options	Setting Up Your Election	4
paper_ballots	Setting Up Your Election	5
before_you_open	Setting Up Your Election	6
election_states	Setting Up Your Election	7
managing_your_voters	Setting Up Your Election	8
emails_to_voters	Setting Up Your Election	9
letting_voters_change_their_vote	Setting Up Your Election	10
polls_and_multiple_races	Setting Up Your Election	11
security_options	Setting Up Your Election	12
choosing_a_tie_breaking_rule	Setting Up Your Election	13
reading_your_results	Counting and Results	1
top_score_vs_winner	Counting and Results	2
preliminary_results	Counting and Results	3
bloc_star	Counting and Results	4
ties	Counting and Results	5
hand_count	Counting and Results	6
verifying_a_tie_break	Counting and Results	7
exporting_your_data	Counting and Results	8
tips_and_tricks	Reference	1
faq	Reference	2
glossary	Reference	3
how_to_enable_beta_features	Reference	4

Reminder that bites: parent: must match the target page's title: character-for-character or the page silently vanishes from the nav — verify with a local build, not by eye.

2. The cross-link wiring PR — READY-AFTER-MERGE (the content PRs)

Pages could only link targets already on main (jekyll-relative-links rewrites only links whose target exists), so cross-references are deliberately thin. The full wiring list — ~52 link insertions across 24 pages, itemized per page — is section 5 of the consistency review. One PR, after the batch lands.

Specific one-liners folded into the same PR:

• election_or_poll.md: restore the two links to election_states.md (removed because #1502 was unmerged).
• is_my_vote_secret.md: restore the After You Vote related-link.
• verifying_a_tie_break.md / choosing_a_tie_breaking_rule.md: link each other.
• using_another_language.md: delete the three "in review" markers as de/fr/it merge (three words each).
• exporting_your_data.md: after #1576 merges, state that overvote_rank/has_duplicate_rank are per-race columns.

3. Duplication trims and terminology alignment — READY-AFTER-MERGE

From the consistency review, sections 2–4: nine duplications (largest: managing_your_voters §"Emailing your voters" reproduces nearly all of emails_to_voters — the latter should own it) and the terminology table (lead item: "organiser" vs "administrator" vs "admin" splits three ways across the voter pages; pick one, sweep once). Best done as one editorial PR after the batch lands, so the diffs are against real files.

4. i18n per-vocabulary keys — AWAITING-DIRECTION (#1574)

Offered upstream: fix the gendered-noun problem for the worst ~12 strings via _election/_poll variants or i18next context. Blocked on maintainers choosing a direction. Evidence and string list are in the issue.

Not all of it is blocked. analysis/polish-translation-fixes.md works the Polish case string by string and finds that six stiff strings and two ungrammatical ones are fixable in pl.yaml alone — the keyword.<election|poll> block is spread as interpolation values and is per-locale, so Polish can carry its own declension forms with no code change and no effect on any other locale. Only this_election_uses genuinely needs the #1574 decision. That queue is READY once the freeze lifts, independently of this item.

5. CI check for untranslated locale values — AWAITING-DIRECTION (#1575)

Offered upstream: ~20-line check (byte-identical-to-English count per locale, ratchet variant). The validation logic already exists — it is the script used to verify all seven locale PRs; porting it to the repo's CI is mechanical once a maintainer says which harness (plain node vs vitest).

6. Quick-poll ownership — AWAITING-DIRECTION (#1556 comment)

Three options posted (own-on-signin / make claim-key work / document fire-and-forget as intended). Options 1 and 3 are small; do not start until answered.

7. Six code fixes — READY (branches committed locally, nothing pushed)

Tabulator and results-page defects, each with a root cause traced to a line, a failing-before test, and a QA page. The first five sit on local branches cut from origin/main 454a38ae, #1480 from b089323f; none is pushed, and no fork branch exists yet. git worktree list inside any BV clone shows the worktrees (bv-1470 is a plain clone, not a worktree).

Issue	What the fix does	Branch @ commit · worktree	Evidence	Page
#1469	Ranked Robin walks its own tiebreak ladder (1st Degree over the finalists, then 2nd Degree over the field) before the random rung. Today a 3+-way tie — i.e. every three-candidate Condorcet cycle — goes straight to the shuffle	fix/1469-ranked-robin-margins-tiebreaker @ 585b08f1 · bv-rr-degrees	jest 52 → 58 green; 4 of the 6 new tests fail on main, 2 regression tests pass both sides; tsc clean	1469
#1507	Allocated Score stops reporting tieBreakType: 'random' when nothing tied — the check was tautological, so every STAR-PR election ever tabulated claimed a random tiebreak, and the results page announced "Tied!"	fix/1507-star-pr-tiebreaktype @ 9a2b8b2a · bv-1507	jest 52 → 54 green; verified against live production bvhchj (7 seats, unique max in all 7 rounds, still reported random)	1507
#1484	One NaN in runBlocTabulator's comparator (-Infinity - -Infinity) stopped its lexicographic sort one key early, so the Race Details tables showed the second-highest scorer instead of the tiebreak runner-up	fix/1484-race-details-runner-up @ a892a0ff · bv-1484	jest 53/56 → 56/56; comparator replayed over the frozen qhjyr2 payload reproduces production's order exactly	1484
#1035	NaN% in the runoff table and the blank runoff pie, when every counted ballot rates both finalists equally. Display guard only — no winner, tally or percentage changes	fix/1035-runoff-zero-denominator @ 47d241a4 · bv-1035	jest 52 → 53 green; probe evaluates the shipped formatPercent rather than a transcription	1035
#1470	The ballot stat tests run on the zero-filled marks, so a ballot scoring every official candidate equally is no longer dropped as an abstention when a write-in is approved — the class of discarded ballots that handed live 43jp39 to the write-in	fix/1470-write-in-abstention-normalization @ c2fc5bd8 · bv-1470	jest: 2 result assertions fail on main, 49/49 green with the fix; full backend suite 179/179; tsc clean; pre-fix production numbers snapshotted in analysis/1470-probe/	1470
#1480	Ranked Robin re-sorts summaryData.candidates into elected-first order — one line, the idiom IRV already uses (sortCandidates(…, 'copelandScore', results.roundResults)) — so the winner its head-to-head rung elects is also the row the frontend highlights. Row order only: elected, every tieBreakOrder value, and perm are byte-identical	fix/1480-ranked-robin-summary-order @ 7d679ba5 · bv-1480	jest: new regression test fails on main at exactly the defect (candidates[0] = the pairwise loser), 6/6 RR suite and 55/55 full tabulator suite green with the fix; tsc clean; first test coverage of the head-to-head rung's log line	page

Before any of these opens as a PR:

1. Rebase on the then-current origin/main and re-run npx jest src/Tabulators/ in packages/backend (setup recipe: npm install, then npm run build in packages/shared, then npm run generate:openapi — without the last two, every tabulator suite fails on a missing module and it looks like the fix broke them).

2. 1469 should first absorb two things from the duplicate branch fix/1469-ranked-robin-margins-tiebreakers @ 709d5c2e (parked in bv-copy-fix): the 69-ballot five-way-cycle regression test, and printing the margin sums in the log lines. Credit that session in the PR body.

3. 1484 and #1035 touch the same results page from different sides — #1484 changes which pair the runoff table calls the finalists, which is what decides whether #1035's denominator is zero. Open them in either order, but say so in both bodies.

4. 1035 introduces two new en.yaml strings that are not approved copy. They need a wording decision, or the PR should propose them explicitly as such.

5. Each page's "could not verify" section is real: none of the three display fixes was rendered in a browser. If the local stack is up when the freeze lifts, run BV2264 first.

6. 1470's change is in the shared filterInitialVotes, so its PR body should carry the blast-radius notes from its page (only STAR / Allocated Score classifications can move; the all-marks-zero methods are invariant under zero-filling). After it deploys: live 43jp39 race 1 must flip to match race 2 (BV2263 is the acceptance check), and #1478 should be re-tested — same root cause if those ballots reach the tabulator as missing keys.

7. #1480 is double-gated: the freeze, and the issue was closed 2026-08-20 as by-design (frontend trusts the backend's order). The fix is the backend half of that very convention, but it opens as a PR only after Arend accepts the reframe — reopen or a fresh backend-scoped ticket (see the closure section of its page). It composes with #1469: disjoint hunks in RankedRobin.ts (the ladder body vs the runBlocTabulator call site), and the re-sort reads whatever winner the ladder produces, so it is correct under either ladder.

---

---

Not parked here, deliberately

• Security-adjacent findings (roll-state enforcement, single-ballot endpoint gating, voter-ID reveal permission). This repo is public; per its own ground rules these go to Arend first, one at a time, framed as "is this deliberate?". They live in the session notes with file:line evidence, ready for that conversation.
• Tabulation cases — they belong in star-voting-library, per the README.