diff --git a/docs/branch-review-ledger.md b/docs/branch-review-ledger.md
index 1af6de2af..b45573806 100644
--- a/docs/branch-review-ledger.md
+++ b/docs/branch-review-ledger.md
@@ -623,4 +623,6 @@ Use this ledger to prevent repeated branch and PR reviews when the reviewed HEAD
 | 2026-07-19 | origin/main 24-hour merged window (d1937d78e..ef042cacd, ~30 PRs incl. #853/#859/#861/#865/#868/#871–#874/#879/#885–#888/#890–#894/#896) | ef042cacd | integrated post-merge regression audit of the full 24h window (code review + design/performance/defect angles) | No P0/P1 regression found at the merged tip. One confirmed P2: the new settings surface (`use-app-preferences.ts` + `settings-dialog.tsx`) presents jurisdiction, population, answer-style, landing, home-content, compact-citations, and all notification preferences as live controls, but no consumer reads them — only density/motion (html attributes + globals.css) and theme are functional; a clinician selecting "Conservative" answer style reasonably but wrongly believes generation changed. One plausible P2/P3: `AuthProvider.initializeSession` now requires a live `getUser()` round-trip, so a transient network failure on load resolves a valid stored session to signed_out (INITIAL_SESSION replay is also skipped); the deliberate stale/tampered-token defense does not distinguish retryable network errors. One P3: `/medications` legacy redirect drops the query string while `/applications` and `/differentials/presentations` preserve theirs. Cleared after inspection: worker image-placement dedupe (key symmetric on both sides), title-word purge/scope migration (matches its replayed review), RPC-layer-only cancellation consolidation (intentional per PR #861), services route-chunk fix intact with no other heavy client value-imports (type-only imports verified), forced-colors ButtonFace/ButtonText flip, codex-autofix/pr-policy workflow changes conform to AGENTS.md (pin retained, rename-aware routing, workflow_sha checkout), audit-metadata allowlist exhaustive over the closed AuditAction union, answer-stream merged abort signals. Environment note (not a repo defect): the session-start hook skips `npm install` when node_modules exists, so this window's dependency bumps left the container stale and `verify:cheap` failed at typecheck until `npm ci`; the hook should compare a lockfile hash. | `npm ci` then typecheck clean and full Vitest 2828 passed / 1 failed / 2 skipped — the sole failure is the long-baselined container-only `tests/pdf-extraction-budget.test.ts` artifact (hosted-CI-green through #826/#835/#872/#890). First `verify:cheap` run passed every static guard (runtime, actions-pin, ci-scope, ci-triage, pr-policy, sitemap, brand, type-scale, icon-scale, function-grants, owner-scope, lint) before the stale-deps typecheck stop. No OpenAI, Supabase, deployment, or provider-backed check ran. |
 | 2026-07-19 | origin/main foreign merges post-#896 landing (541da7b #871 remediation + f4557ca #892 policy parsing; explicit user review request) | ef042ca6a34d33862936e77e4b71068978382c1e | Bug review of recent main changes (diff-review protocol) | One P2 confirmed and fixed in PR #905: #892's widened heading matcher ended a required PR-body section at ANY next heading, so `###` sub-structure inside `## Verification` truncated the section and false-rejected valid bodies (fail-closed; repo template unaffected; repro via direct evaluatePullRequestPolicy probe old-vs-new). No P0/P1. Cleared after verification: `src/lib/client-env.ts` (no env leakage; production demote-to-demo removal deliberate and fail-safe — upload gating still locked via canUsePrivateApis), applications/medications redirect routes (fixed targets, 307+HEAD alias, no open-redirect/header-injection), test infra (no weakened assertions; route-coverage spec added to all projects), and the three biggest #871 UI diffs read inline (ClinicalDashboard upload tablist roving-tabindex + hydration-safe useSyncExternalStore role switch; visual-evidence unavailable-source rows became real non-interactive elements; favourites library demo-gates prototype items with sound menu keyboard nav). Residual (report-only): pr-policy `section()` remains fence-unaware (headings inside fenced code can satisfy required-section checks — pre-existing class, author-controlled attestation surface); dev-only Turbopack persistent-cache staleness served an old globals.css compile across restarts twice this session (fixed by setting `.next` aside). | Reviewer fan-out: general lane completed with concrete probes (pr-policy self-test, node repro on old parser from f4557ca^, adversarial body probes); UI lane agent lost to session limit and re-done inline on the three biggest diffs, leaning on the merged tree's green gates (verify:ui 236-passed run in this session covers #871's own new specs). No provider-backed checks run. |
 | 2026-07-19 | claude/clinical-kb-pwa-review-asi3wb (PR #905; commits b2afe66 visuals + 6531178 policy + this ledger follow-up) | 6531178c1a3427dfd58c4bfcf8e29020c5731179 | PWA install/update notice redesign (all breakpoints) + review-follow-up policy fix | Redesigned the five PWA notices (install, update, iOS hint, offline, restored) as glass lux cards: per-type semantic icon tiles, heading-ink titles, corner dismiss buttons, reduced-motion-safe 280ms entrance. Deliberate placement per screen size: phones keep the bottom card above the fixed composer (thumb zone, safe areas); ≥640px floats a 25rem card bottom-right; ≥1280px moves the stack to a top-right toast under the header (same 4.25rem+safe-area offset constant as the mode-menu popover) with the animation direction flipped. Copy, roles, and button names unchanged. Plus the outline-aware pr-policy section parser fix from the same-session review (see the review row above). | Focused vitest pwa-lifecycle.dom 9/9 + pwa-manifest 8/8; `check:pr-policy` self-test green incl. new sub-heading case; `verify:cheap` 2828/2831 (sole fail = known container-only pdf-extraction-budget artifact); `test:e2e:pwa` privacy/offline green (installability fail = known container `in-incognito` artifact); `verify:ui` 236 passed/2 failed (the two long-baselined container artifacts); production build + client-bundle secret scan + bundle budget within tolerance (1293.4 vs 1278.6 KiB baseline); visual evidence at 390/768/1440 light+dark+offline in session scratchpad. Dev caveat recorded: Turbopack persistent `.next` cache served stale globals.css across restarts twice; fixed by setting the cache aside. No provider-backed checks run. |
+| 2026-07-19 | `origin/main` through PR #903 plus fixed 48-hour PR snapshot (`#689`–`#902`) | 4034d2e60ebb6616130ff17bf3cb69368f36f8f6 | whole-repository, all-lens regression and PR-activity review | Changes requested. No P0. Confirmed five P1 defects: stale publication approvals are not bound to reviewed document state; invalid supplied credentials can become anonymous uploads; pooled duplicate uploads expose another uploader's metadata; readiness is fail-open for database usability errors; and settings promise clinical tailoring/alerts with no consumers. Eighteen P2 findings cover semantic rerank effectiveness/privacy, summary prompt trust, PDF resource limits, cancellation, auth-state loss, public storage-path exposure, query validation, provider-boundary/CI/test gaps, factsheet/Therapy behavior, Therapy startup cost, and the PR #903 ledger's incorrect claim that PR #901 has zero unresolved threads. GitHub GraphQL still reports two current unresolved P2 threads on merged PR #901. | Exact tree `68a58f6f..4034d2e60`: 217 commits, 566 files, +59,742/-8,242. Fixed snapshot inventory: 213 PRs created and 25 older PRs updated. On `a871dd765`, `verify:cheap` passed (317 files/2,879 tests), offline RAG passed (21 suites/294 tests), and production build plus required Chromium passed (1,682 pages; 239/239). PR #899 exact head `8242fa63d` has the same full local proof and green hosted checks; PR #903 is docs-only and passed `git diff --check`, docs links, and docs script references. Production-readiness CI, design-system, env parity, workflow guard, and offline audit passed. `docs:check-index` remains advisory-red and full-range `git diff --check` reports four intentional Markdown hard breaks plus three SQL whitespace lines. No OpenAI, Supabase, deployment, live clinical, or provider-backed release command ran. |
+| 2026-07-19 | local branches and worktrees after PRs #905 and #907 | e377ab1aed44d56c303eede80572c1df82ddcd6e | final local branch/worktree cleanup and useful-history preservation | Reduced 153 local branches to `main` plus the two actively edited task branches. Deleted 150 redundant refs using cherry-pick containment, exact merged-PR heads, synthetic no-op merge trees, exact duplicate-head retention, merged-PR commit association, and explicit supersession review. Removed two clean obsolete worktrees and retained only the active auth/Supabase and P2-remediation worktrees for their owners. Archived the final 66 superseded historical heads in a verified 76,395,256-byte Git bundle before ref deletion. PRs #905 and #907 merged through protected `main`; PR #906 subsequently merged through protected `main` and its remote branch was removed. | Fresh fetch/prune and PR inventory; exact-old-value `git update-ref`; `git merge-tree --write-tree`; GraphQL commit-to-PR association for 178 commits; reverse-patch checks against a detached fresh-main worktree; verified bundle `Database-local-refs-before-final-cleanup-20260719-0525.bundle`; final branch/worktree/remote checks. No OpenAI, Supabase, deployment, live clinical, or production-data workflow ran. |
 | 2026-07-19 | claude/audit-recent-changes-kde66i (audit-remediation follow-up to the ef042cacd audit row) | see PR head | remediation of the three 2026-07-19 audit findings | Fixed all three findings from the 24-hour merged-window audit. (1) Inert settings honesty (P2): every preference the app does not yet consume — jurisdiction, population, answer style, landing, both home-content toggles, compact citations, and all three notification toggles — now renders an explicit "Saved for later — not active yet" marker in `settings-dialog.tsx`; the functional appearance/density/motion rows stay unmarked. Wiring the preferences into answer generation was deliberately NOT done here (clinical-behavior change requiring governance review); the new dom test documents the contract for flipping a control live. (2) Auth offline resolution (P2/P3): `resolveInitialAuthState` gains `verificationUnavailable`, set from `isAuthRetryableFetchError(getUser().error)`, so an unreachable auth server keeps the stored session signed in while a reachable server that rejects the token still resolves signed_out; server-side bearer validation is unchanged. (3) `/medications` redirect (P3): now preserves the sanitized q/focus/run search context with the same allowlist as the root legacy-mode redirect. | Focused Vitest 30/30 across `settings-inert-preferences.dom` (new), `private-client-auth` (3 new cases), `audit-navigation-auth-regressions` (query-preservation case), `app-preferences`, and `site-map`. Full `verify:cheap` run recorded on the PR. No OpenAI, Supabase, deployment, or provider-backed check ran. |