Docs coherence pass: fix stale status/cross-refs, add index + orientation

[WilfordGrimley]

Note: this branch also carries Proposal G's earlier survey/spec commits
(user accounts + saved decks, already reported separately) — that work is
a HOLD, no build started; this PR's own scope is the docs coherence pass
below only.

Description

Survey-then-fix documentation coherence pass across docs/ after five
days of multi-worker sprint activity. Mechanical fixes only — no feature
doc's technical content was rewritten, and docs/theory.md's substance
was not touched.

Inventory / survey (four parallel agents, one per doc cluster, plus
manual spot-checks): every doc in docs/ was read in full and checked for
(1) stale status/date claims contradicted by later content in the same
file or by the actual codebase/git history, (2) broken cross-references,
(3) orphaned docs nothing points to. Findings below; anything that needed
judgment rather than a mechanical correction is listed as "not fixed,
flagged" rather than acted on.

Fixed:

• catalog-completion-plan.md — 3 of 6 ## Part N headers (1, 3, 6)
  carried stale pre-condition parentheticals ("in progress", "after PR Hash-at-ingest + two-threshold clustering: dedup as a free DB read #27
  merges", "starts only after...") contradicted by the file's own ## Status section; updated to match reality (merged/built/merged).
• printing-tags.md — bumped a 2-day-stale date stamp; removed a false
  "this document's current stage" claim (Stage 6) superseded by Stage 7/8
  content later in the same file; corrected a "full-catalog run: not yet
  fired" claim contradicted by that file's own later sections; disambiguated
  a duplicate pair of "Key files"/"Known gaps" H2 headings left over from
  the pre-rewrite changelog tail (anchor collision hazard).
• proposal-f-public-stats-page.md — updated its queue-order line and a
  since-completed backfill citation, both superseded by proposal-g's later
  decisions/status.
• upstreaming/vote-system.md — repointed a dangling reference to a fact
  no longer recorded in CLAUDE.md.
• troubleshooting.md — an entry marked "still open" (swallowed
  vote-error messages) was actually fixed the same day by PR Editor mobile scroll-chain fix + vote error handling (surfaced messages, friendly 429) #47; updated.
• infrastructure.md — corrected a Postgres/ES port-binding claim that's
  only true for local dev; prod publishes no host ports at all
  (ports: [] override).
• card-dom-api.md, google-drive-connect.md, pdf-generator.md,
  print-export-page.md — stale file paths after component relocations
  into feature folders, plus one reference to a deleted component
  (PrintingTagQueue.tsx, superseded by QuestionFeed.tsx).
• print-export-page.md — added the PringlePrints-flow TODO parity bullet
  missing from Known Gaps (NotMPC's identical TODO was already listed).
• moderation.md — moved a completed/verified item out of "Known gaps",
  and repointed a cross-reference to where its content actually lives (a
  code comment on post_submit_printing_tag, not printing-tags.md or
  troubleshooting.md as previously miscited).
• CLAUDE.md — added the missing docs/features/moderation.md index
  entry (every other feature doc was already listed), and a pointer to the
  two new docs below.

Added:

• docs/README.md — audience-grouped map (understanding the system /
  operating it / plans & proposals / records), one-word status per
  proposal doc.
• docs/overview.md — orientation page for a cold external reader. The
  federation pitch sends the upstream maintainer to docs/theory.md; this
  is what greets them if they look further: what this fork is, its
  relation to upstream, and the three big systems (catalog + votes,
  identification pipeline, print tooling).
• docs/reports/README.md — scaffolds the report-relay convention's
  directory ahead of the report-relay/report-relay-2 branches merging
  (per owner confirmation mid-task: the convention is new, its files live
  on those branches, and merge in a docs batch soon).

Not fixed — flagged for the owner instead:

• Several "as of this writing" PR-status claims in infrastructure.md's
  upstreaming section (PR Measure STAGE_E_MICRO_BATCH_SIZE at the next real bulk inflow (25/50/100 protocol) #463/Stage E shakedown driver: route the Bug-A tail through the streaming conveyor #465 open/closed status, an upstream
  reviewer's stated intentions, a GitHub Support ticket's status) — can't
  be verified live from this sandbox (no network access to GitHub), so left
  as-is rather than guessed at.
• Dated, changelog-style narrative bled into otherwise-durable
  lessons.md/troubleshooting.md entries (e.g. specific incident dates
  and one-off measurements inside what's meant to be terse, reusable
  reference) — a documentation-convention question, not a factual error,
  so left for the owner's call rather than trimmed unilaterally.
• docs/audits/ui-content-audit.md exists (12 UI content-accuracy
  findings, HOLD) but only on the unmerged claude/ui-content-audit
  branch — that PR (UI content-accuracy audit findings (docs, HOLD) #56) is deliberately held until the audit worker adds
  a per-row disposition column (PRs Build pass on UI content-accuracy audit selections (PR #56 items 1-10, 12) #64/GIS script-load failure: actionable Drive-save error (lazy-load audit: already correct) #65). Noted in docs/README.md's
  Records section, not merged here.
• docs/reports/'s actual content lives on report-relay/report-relay-2,
  unmerged — only the directory's orientation README was added here, per
  the owner's explicit instruction to write it ahead of that merge.

Checklist

• I have installed pre-commit and installed the hooks with pre-commit install before creating any commits.
• I have updated any related tests for code I modified or added new tests where appropriate.
• I have manually tested my changes as follows:
  • Verified every corrected file-path/component claim against the actual frontend/src tree (find/grep) before editing, rather than trusting agent claims uncritically — caught and corrected one agent-reported cross-reference that turned out to be wrong (moderation.md's rate-limiter caveat isn't actually documented in troubleshooting.md, only in a views.py code comment).
  • Verified the Postgres/ES port-binding fix against the actual docker/docker-compose.prod.yml contents.
  • Ran npx prettier@2.7.1 --check across every touched Markdown file plus the two new ones; all pass.
• I have updated any relevant documentation or created new documentation where appropriate.

---

Generated by Claude Code