diff --git a/.env.example b/.env.example
index 5bc02df9c..1949c48f4 100644
--- a/.env.example
+++ b/.env.example
@@ -129,18 +129,20 @@ AISSTREAM_API_KEY=
 # 0 disables the silence watch entirely.
 # AISSTREAM_SILENCE_TIMEOUT_MS=120000
 
-# Optional: TomTom live traffic flow (BYOK freemium — free tier is ~50,000 tile
-# requests/day, get a key at https://developer.tomtom.com). SERVER-SIDE ONLY:
+# Optional: TomTom live traffic flow (BYOK freemium — the free tier is 200,000
+# Traffic Flow & Incidents tile requests per MONTH, not per day; no credit card
+# required. Get a key at https://developer.tomtom.com). SERVER-SIDE ONLY:
 # the browser fetches same-origin /api/tomtom/* and never sees the key.
 # Keyless fallback: the traffic layer runs its built-in simulation (white dots,
 # hardcoded per-road-class speeds) — no key required for the layer to work.
 # TOMTOM_API_KEY=
-# Optional soft cap on upstream tile fetches per UTC day (default 40000) — a
+# Optional soft cap on upstream tile fetches per UTC day (default 6000) — a
 # configurable application safety ceiling, not a guarantee of staying within
 # TomTom's free allowance (currently 200K tile requests/month:
-# https://docs.tomtom.com/pricing/). Over the cap the proxy serves cached/stale
-# tiles instead of hitting upstream.
-# TOMTOM_DAILY_TILE_BUDGET=40000
+# https://docs.tomtom.com/pricing/). 6000/day keeps a 31-day month at 186K,
+# inside that allowance. Over the cap the proxy serves cached/stale tiles
+# instead of hitting upstream.
+# TOMTOM_DAILY_TILE_BUDGET=6000
 
 # Optional: CCTV layer tuning (advanced). Defaults are sensible — leave unset
 # unless you're customizing the camera source pack.
diff --git a/.gitignore b/.gitignore
index 491a82c15..1a64fff39 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,6 +1,13 @@
 node_modules
 dist
+# Credential files. `.env` alone is not enough: Vite's loadEnv (see
+# server/standalone/vite.config.js) and scripts/setup-doctor.mjs both read the
+# full dotenv ladder — .env, .env.local, .env.<mode> and .env.<mode>.local —
+# so a key placed in any of them is live for the app. Only .env was ignored,
+# leaving the other three committable. .env.example is the tracked template.
 .env
+.env.*
+!.env.example
 *.log
 screenshots/
 qa-shots/
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 0f3fd7822..da600f5b0 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,30 @@
 # Changelog
 
+- Report AIS speed and course that carry the standard "not available" code as
+  unknown instead of 102.3 knots and 360 degrees. Genuine readings, including a
+  stopped vessel's zero and the highest encodable values, are unchanged.
+
+- Render native `<select>` option lists in the dark UI palette. The closed
+  controls were already skinned, but the browser-painted popups fell back to the
+  platform light palette, leaving near-white option text on a white surface in
+  the HUD layout, Scenes, CCTV camera, Radio filter and Draw colour menus.
+
+- Credit adsbdb, which supplies the aircraft type, model name and registration
+  on enriched flights and the airline and origin/destination pair behind the
+  tracked contact's route strip. `DATA_SOURCES.md` now records adsbdb's
+  published credits and route-data restriction, along with the request bounds
+  and gitignored 24-hour local cache. A matching `DATA_CREDITS` entry surfaces
+  the credit in the in-app Data attribution popover, and a test protects it
+  against accidental removal.
+
+- Size the TomTom daily tile budget to the provider's real free allowance.
+  `TOMTOM_DAILY_TILE_BUDGET` defaulted to 40,000/day against an allowance
+  granted monthly (200,000 tile requests/month), exhausting a month in five
+  days and leaving the traffic layer dead for the rest of the period. The
+  default is now 6,000/day (186,000 over a 31-day month). Corrects the stale
+  "~50k/day" free-tier figure in the proxy, `.env.example` and
+  `DATA_SOURCES.md`. Still an application-side ceiling, not a billing cap.
+
 - Distinguish PARTIAL vessel snapshots from STALE data in the layer panel, with
   accepted-record counts and unchanged retention, freshness and outage safeguards.
 
@@ -108,6 +133,10 @@
 
 - Expose portable radio, camera-type and regional source helpers; keep HTTP transport separate from record normalization.
 
+- The Realtime debug-log endpoint is bounded on every axis it was not: an always-on per-client rate limit, asynchronous appends through a serialized queue instead of a synchronous write on the request path, and rotation of the log file at 32 MB keeping one prior generation. The 8 MB cap applied to a single request body and never to the file those requests accumulated into, so any local page could grow it for as long as the dev server ran. A malformed record and a failed write are now told apart, 400 from 500, and neither answer carries the error text.
+
+- Three proxy paths no longer relay upstream or JS error text to the client. The HUD summary passed OpenAI's own `error.message` through whenever upstream was not ok, carrying request ids and quota wording; the Realtime token route passed through non-success response bodies and echoed JS errors, which can expose upstream details; and a failed CCTV media fetch stored the raw errno as the camera's health message, which reaches the screen through `GET /api/cctv/health` rather than through the sanitized response beside it. Logs now name the failure and the upstream status without the text.
+
 - Separate vessel records and feed acquisition from rendering while preserving selection, partial-feed retention, sea-surface placement and request cancellation.
 
 - Separate military-flight records and acquisition from rendering while preserving ground-model ownership, source units and follow behavior.
@@ -242,6 +271,15 @@ of current runtime behavior, see [`docs/CURRENT-STATE.md`](docs/CURRENT-STATE.md
 
 ### Fixed
 
+- Keep traffic-road bounds crossing the antimeridian monotonic and inside the
+  longitude range accepted by the Overpass request path, preserving the small
+  wrapped span instead of producing an inverted or rejected box (#392 — thanks
+  @Ashfaqbs).
+- Make `npm run doctor` report keyless anonymous OpenSky access for explicit
+  `OPENSKY_AUTH_MODE=anon` and OAuth mode without a client pair, retain the
+  existing OAuth-pair capability wording, and identify selected Basic or auto
+  modes without guessing their eventual credential choice. OpenSky proxy
+  authentication is unchanged.
 - Bikeshare stations load again. The extracted station source addressed the
   proxy as `/api/gbfs?url=`, but the proxy reads its upstream target from the
   path, so every request answered 400 and the layer reported a fetch error for
diff --git a/DATA_SOURCES.md b/DATA_SOURCES.md
index 850d27e39..7c822558a 100644
--- a/DATA_SOURCES.md
+++ b/DATA_SOURCES.md
@@ -20,6 +20,7 @@ How to read this:
 | **OpenSky Network**                                                   | Primary worldwide live-flight snapshot                                                                                              | Non-commercial research/education license                                                                                                                                                                                                                                                                                                             | Schäfer et al., _"Bringing Up OpenSky"_, IPSN 2014 + opensky-network.org                                                                    |
 | **adsb.lol point API**                                                | Bounded live-flight fallback when OpenSky has no usable snapshot                                                                    | ODbL 1.0                                                                                                                                                                                                                                                                                                                                              | adsb.lol contributors; `api.adsb.lol/v2/lat/{lat}/lon/{lon}/dist/{radius}`                                                                  |
 | **adsb.lol**                                                          | Military flights + aircraft traces                                                                                                  | ODbL 1.0                                                                                                                                                                                                                                                                                                                                              | "adsb.lol" (ODbL)                                                                                                                           |
+| **adsbdb** (`api.adsbdb.com`) | Aircraft type, model name and registration by ICAO hex for enriched flights; airline and origin/destination airports for the tracked flight's callsign | No licence or rate limit is published for the service. Its own credits: aircraft data from PlaneBase, ICAO-to-N-number conversion by Guillaume Michel, and flight route data that "is the work of David Taylor, Edinburgh and Jim Mason, Glasgow, and may not be copied, published, or incorporated into other databases without the explicit permission of David J Taylor, Edinburgh" | "adsbdb — api.adsbdb.com" (courtesy); route data credited to David Taylor, Edinburgh, and Jim Mason, Glasgow |
 | **AISStream.io**                                                      | Live vessels (AIS)                                                                                                                  | Free, beta, no formal ToS; AIS is a public broadcast                                                                                                                                                                                                                                                                                                  | "AISStream.io" (courtesy)                                                                                                                   |
 | **CelesTrak**                                                         | Satellite TLEs (SGP4)                                                                                                               | US-government-origin data, no license; citation requested                                                                                                                                                                                                                                                                                             | "CelesTrak (celestrak.org), Dr. T.S. Kelso"                                                                                                 |
 | **The Space Devs — Launch Library 2 v2.3**                            | Recent launch, payload, stage, and recovery metadata for Space Missions (30d)                                                       | [The Space Devs terms of use](https://github.com/TheSpaceDevs/Tutorials/blob/main/faqs/faq_TSD.md#terms-of-use): data may be used and shared in any form; avoid forwarding it without added value; attribution is encouraged (not mandatory). [Official API limits](https://ll.thespacedevs.com/docs/): 15 unauthenticated calls/hour; optional token | "Launch Library 2 — The Space Devs" (courtesy attribution)                                                                                  |
@@ -62,6 +63,7 @@ How to read this:
 - **Google Maps Platform.** You supply your own API key and are bound by [Google's ToS](https://cloud.google.com/maps-platform/terms). Google Maps Content (tiles, geocodes, places) **may not be cached, stored, rehosted, or committed** — this app only ever uses it live, which is the compliant pattern. The "Google" attribution is displayed on the globe and must stay visible. Restrict your key (see [SECURITY.md](SECURITY.md)).
 - **OpenSky Network.** Its license is **non-commercial**, and operational use of the REST API in a live product can require a prior written agreement with OpenSky — even for non-profit/government use. If you deploy this commercially, contact OpenSky for your own terms. The flights layer is a toggle and runs anonymously by default.
 - **adsb.lol flight fallback.** When OpenSky is unavailable and no last-good OpenSky response exists, the server requests a cached, capped 250 nm adsb.lol point snapshot around the current camera subpoint. This is regional observed context, not worldwide completeness; provenance is exposed in the Flights stats/context row. Military ICAOs remain reconciled through the existing dedicated military registry rather than duplicated.
+- **adsbdb enrichment.** `/api/adsbdb` resolves an ICAO hex to aircraft type, model name and registration, and the tracked flight's callsign to an airline and an origin/destination pair. Fan-out is deliberately small because adsbdb is a free community API that publishes no quota: at most 4 requests in flight, dispatches dripped to roughly 5/s, and each key asked for at most once per session, with route lookups reserved for the tracked aircraft. The dev-server proxy caches each key for 24 hours in `.gev-cache/adsbdb.json`, negative answers included, so fresh entries survive server restarts. That gitignored cache remains local to the deployment and is not bundled with this repo. adsbdb publishes no general service license; its README says the route data may not be copied, published, or incorporated into other databases without the explicit permission of David J Taylor, Edinburgh. Deployers are responsible for obtaining any permission their use requires. Enrichment is best-effort and fail-silent — a flight that gets no answer keeps what the flight feed already gave it.
 - **Launch Library 2.** `/api/launches` makes a server-side rolling-30-day query against the supported v2.3 detailed launch endpoint, caches successful responses for 15 minutes in memory and on disk, and serves the last successful response during a throttle or transient outage. Anonymous access is limited to 15 calls/hour; deployments can provide `LL2_API_TOKEN` for authenticated access. The Space Devs' published terms permit using and sharing the API data in any form, ask users not to forward it without adding value, disclaim complete accuracy, and encourage—but do not require—attribution. This app keeps a courtesy credit. Payload and stage/recovery records are shown only when supplied. Failed launches expose their source status and never receive fallback orbit geometry or a live/estimated marker. LL2 supplies launch context and event timing, not continuous ascent telemetry or live orbital state.
 - **TfL JamCams.** The camera list comes from the keyless `api.tfl.gov.uk` endpoint (an optional `TFL_APP_KEY` raises its rate limit); frames come from TfL's public S3 bucket. The "Powered by TfL Open Data" attribution is required by TfL's terms and is registered in the Data attribution popover.
 - **Ontario 511 CCTV.** The camera list comes from the keyless `511on.ca/api/v2/get/cameras` endpoint; frames are server-registered still images under `511on.ca/map/Cctv/`. The source is live-fetched at runtime and attributed under the Open Government Licence - Ontario.
@@ -76,7 +78,7 @@ How to read this:
 - **Camera frame content.** Public camera frames can contain people, vehicles and license plates, and the camera layer does nothing with them beyond putting them on the map. A successful upstream response is relayed as the provider served it: nothing in the camera pipeline blurs, redacts, enhances or restores it, and nothing there runs face, plate or object detection or otherwise recognises or classifies what is in the picture. (The ALPR layer described below maps where plate-reader cameras are, from locations contributors tagged in OpenStreetMap; it reads nothing from any camera frame.) When an upstream has no frame for a camera, what you see instead is a Google Street View still or the app's own placeholder card; the panel's source badge names which of them you are looking at, and a refresh that fails after a frame has arrived leaves the last good picture on screen rather than replacing it. The browser resamples each frame for display: onto a fixed 1920x1080 surface for the camera's projection plane in the scene, and onto a 192x108 canvas for the small ambient card. That scaling is the only change made to the picture, and the only pixels the camera layer reads back are a 64x36 downscale hashed to tell whether the picture has changed. Live streams are piped through; the still path reads the image into memory to pass it on — TxDOT wraps its JPEG in a JSON envelope, which is decoded there — and no frame is written to disk (the disk cache, `.gev-cache/`, holds feed JSON for other layers and never camera imagery). Outside that pipeline, the voice assistant is the one feature that sends imagery anywhere: when it answers a question about what is in view it may attach a downscaled screenshot of the rendered scene — including any camera plane on screen — to the request it sends to the voice model. Frames are fetched only from server-registered camera URLs — a client cannot hand the proxy a URL (see [SECURITY.md](SECURITY.md)). What the app does stays inside the line README draws: no named-person search, no face recognition, no tracking individuals.
 - **Transit (GTFS-Realtime).** The proxy requests only registered operator feeds, validates redirects, bounds responses, and uses conditional requests, short snapshot caching and failure backoff. All seven registered regions are live-enabled; each can be disabled independently. Vehicles use delayed playback between reports, with report age and waiting states shown separately. Session-local selected trails are available for every feed. MBTA additionally retains up to 15 minutes of recently observed positions in the running proxy, so available history can survive a browser reload; server restart clears it. History reads never fetch upstream, and other feeds have no proxy retention. Nothing is archived to disk or collected in the background. Operator text credits are included as a courtesy and do not imply endorsement. Heights are aligned to work with Google 3D tiles; subway markers are shown at street level because these feeds do not supply depth.
 - **Radio Browser.** `/api/radio/stations` discovers official API mirrors, makes bounded and coalesced healthy/geolocated HTTPS-station queries, caches the normalized public-domain directory for 45 minutes, and may serve the last good catalog for up to seven days during an outage. Refreshes must meet minimum accepted-query and station coverage before replacing a warm catalog; schema-valid responses whose rows all fail the product's health policy do not count as successful queries. A usable partial cold catalog is explicitly `DEGRADED`, and malformed or empty successful payloads are rejected atomically. Every directory and click-count request rejects redirects, validates all resolved addresses as globally routable (including reserved/documentation IPv4 and special/non-global IPv6 exclusions), and pins the TLS connection to a validated address. Only MP3/AAC non-HLS directory rows with public HTTPS stream targets are returned; favicons are intentionally omitted. Pressing play connects one browser audio element directly to the selected broadcaster and calls the directory's click counter through known-ID-only `POST /api/radio/click/:uuid`. GEV never proxies, caches, records, bundles, or redistributes audio. Radio Browser supplies station-level tags, not dependable current-song or upcoming-program metadata, so Radio filtering never claims either. Direct playback exposes the listener's IP address to the broadcaster, whose own stream terms apply.
-- **TomTom Traffic.** Optional and BYOK: without `TOMTOM_API_KEY` the traffic layer runs its built-in simulation and no TomTom data (or attribution) appears. With a key, flow vector tiles are fetched through the server-side `/api/tomtom` proxy (120 s cache + a daily tile-budget governor — `TOMTOM_DAILY_TILE_BUDGET`, default 40,000, a configurable application safety ceiling, not a guarantee of staying within TomTom's monthly free allowance; TomTom's [current pricing](https://docs.tomtom.com/pricing/) lists 200K free tile requests per month) and the "Traffic flow data © TomTom" credit is registered in the Data attribution popover the moment live mode activates. TomTom data is served live and cached only transiently (≤120 s TTL under `.gev-cache/`, gitignored) — it is not bundled or redistributed. One 23 KB point-in-time tile snapshot is committed as a decode-test fixture (`src/data/fixtures/`, © TomTom, never served to the app).
+- **TomTom Traffic.** Optional and BYOK: without `TOMTOM_API_KEY` the traffic layer runs its built-in simulation and no TomTom data (or attribution) appears. With a key, flow vector tiles are fetched through the server-side `/api/tomtom` proxy (120 s cache + a daily tile-budget governor — `TOMTOM_DAILY_TILE_BUDGET`, default 6,000 — sized so a 31-day month stays inside TomTom's 200K monthly free allowance; still a configurable application safety ceiling, not a guarantee; TomTom's [current pricing](https://docs.tomtom.com/pricing/) lists 200K free tile requests per month) and the "Traffic flow data © TomTom" credit is registered in the Data attribution popover the moment live mode activates. TomTom data is served live and cached only transiently (≤120 s TTL under `.gev-cache/`, gitignored) — it is not bundled or redistributed. One 23 KB point-in-time tile snapshot is committed as a decode-test fixture (`src/data/fixtures/`, © TomTom, never served to the app).
 - **Routing (OSRM / FOSSGIS).** `/api/route` proxies the public OSRM instances FOSSGIS e.V. runs for the OpenStreetMap community (car, bike, foot profiles). They are keyless and intended for fair, interactive use — the proxy rate-limits clients (60/min, 200/min total), caps route span and response size, caches each route for 10 minutes, and identifies itself in the User-Agent. Heavy or automated use must run its own OSRM. Routes are computed from OpenStreetMap data (ODbL); the Directions layer and voice routes never redistribute them. A route the router cannot find is reported as such; the Directions layer draws nothing in that case, and voice routes fall back to a clearly labeled straight line.
 - **Re:Earth Terrain.** Keyless (no API key). Used two ways: (1) `src/mapStackController.js` swaps in a `Cesium.CesiumTerrainProvider` pointed at Re:Earth's `cesium-mesh/ellipsoid` quantized-mesh endpoint for globe stacks without a Cesium ion token (e.g. OSM), replacing a flat `EllipsoidTerrainProvider`; falls back to the flat provider if the endpoint can't be reached. (2) The server-side `/api/terrain/heights` proxy (disk-cached, serve-stale) resolves per-point ellipsoidal ground height for entity placement. Both are best-effort with a keyless-safe fallback (bundled EGM96 geoid math) if Re:Earth is unreachable.
 - **OSRM on the FOSSGIS routing servers.** Keyless (no API key). Every route is requested by a person: clicking A and B in the Directions layer, or a voice "route from A to B". The server-side `/api/route` proxy sends an identifying `User-Agent`, pins the upstream host, accepts 2-12 coordinates, rejects any leg over 600 km or a total over 2,500 km, caps the buffered response at 8 MB, times the upstream out at 12 s, rate-limits per client, caches each route for 10 minutes, and coalesces identical requests that are in flight at the same time. Turn-by-turn maneuvers are asked for only when a caller wants them (`steps=1`), so the callers that do not read them send the smaller request. The "fix the map" link the policy asks for is in the Data attribution popover beside the OpenStreetMap credit.
@@ -191,4 +193,4 @@ Douglas-Peucker simplification, 6-decimal rounding).
 
 ## In-app attribution
 
-The required Google Maps / Cesium credit renders on the on-globe credit line (`#cesium-credits`, bottom-left) and must stay visible — including in clean-view and recording modes (the whole line, logo + "Google Maps" + the "Data attribution" link, stays on screen; only the GEV panels/HUD fade). The layer-specific credits (adsb.lol, TeleGeography, OSM datacenters/dams/roads, NASA FIRMS, CelesTrak, USGS, City of Austin, Fintraffic, GBFS, Radio Browser, OpenSky, AISStream) are registered into the expandable **"Data attribution"** popover on that credit line via `viewer.creditDisplay.addStaticCredit(new Cesium.Credit(html, /* showOnScreen */ false))` — see `src/data/dataCredits.js`. When you add a new data source, add its license and attribution to this file **and** append an entry to `DATA_CREDITS` in `src/data/dataCredits.js` so it surfaces in the app.
+The required Google Maps / Cesium credit renders on the on-globe credit line (`#cesium-credits`, bottom-left) and must stay visible — including in clean-view and recording modes (the whole line, logo + "Google Maps" + the "Data attribution" link, stays on screen; only the GEV panels/HUD fade). The layer-specific credits (adsb.lol, adsbdb, TeleGeography, OSM datacenters/dams/roads, NASA FIRMS, CelesTrak, USGS, City of Austin, Fintraffic, GBFS, Radio Browser, OpenSky, AISStream) are registered into the expandable **"Data attribution"** popover on that credit line via `viewer.creditDisplay.addStaticCredit(new Cesium.Credit(html, /* showOnScreen */ false))` — see `src/data/dataCredits.js`. When you add a new data source, add its license and attribution to this file **and** append an entry to `DATA_CREDITS` in `src/data/dataCredits.js` so it surfaces in the app.
diff --git a/README.md b/README.md
index 4224bacc0..580bb147c 100644
--- a/README.md
+++ b/README.md
@@ -34,7 +34,7 @@ _“pretty cool”_ — [Brendan Eich](https://x.com/BrendanEich/status/20945920
 
 <div align="center">
 
-**[Quick Start](#-quick-start) · [First Five Minutes](#-the-first-five-minutes) · [Talk to It](#-talk-to-it) · [What's Live](#-whats-on-the-globe) · [Under the Hood](#-under-the-hood) · [Keys & Costs](#-api-keys)**
+**[Quick Start](#-quick-start) · [First Five Minutes](#-the-first-five-minutes) · [Talk to It](#-talk-to-it) · [What's Live](#-whats-on-the-globe) · [Under the Hood](#-under-the-hood) · [Keys & Costs](#-api-keys) · [Contributing](CONTRIBUTING.md)**
 
 </div>
 
@@ -414,8 +414,10 @@ Add these if you need higher polling allowances.
 where each configured provider was found without printing credential values.
 On macOS its Keychain-aware result previews `./scripts/dev-fresh.sh`; plain
 `npm run dev` reads only explicit environment and Vite dotenv values. The
-OpenSky summary reports only OAuth client-pair presence, not the resolved
-runtime mode or credential validity; Basic and credentials-file modes remain
+OpenSky summary reports keyless anonymous access for explicit `anon` or an
+OAuth mode without a client pair, retains presence-only wording for a complete
+OAuth pair, and identifies selected Basic or auto mode without guessing which
+credentials runtime will accept. Basic and credentials-file modes remain
 advanced `dev-fresh.sh` configuration.
 
 <details>
diff --git a/docs/CURRENT-STATE.md b/docs/CURRENT-STATE.md
index b5727f353..791755972 100644
--- a/docs/CURRENT-STATE.md
+++ b/docs/CURRENT-STATE.md
@@ -1,5 +1,19 @@
 # God's Eye View Current State
 
+AIS encodes speed over ground in 0.1-knot units and course over ground in
+0.1-degree units, reserving the top code of each field for "not available", so
+those reports arrive as 102.3 knots and 360 degrees. Both are stored as unknown,
+and vessel cards, the HUD and analyst queries show a missing measurement rather
+than a reading. Heading keeps its existing separate check for its own sentinel.
+Genuine readings are unchanged, including a stopped vessel's zero and the
+highest speed and course the fields can encode.
+
+Native `<select>` controls declare a dark colour scheme and explicit option
+colours, so the browser-painted option list matches the panel it drops out of.
+The scheme is declared on the controls rather than on the document root, so
+hand-tuned scrollbar and text-input skins are unchanged. Applies to the HUD
+layout, Scenes, CCTV camera, Radio filter and Draw colour menus.
+
 Vessel snapshot completeness is separate from freshness. A current snapshot with
 rejected or duplicate records shows PARTIAL with accepted/received counts; stale
 or unknown freshness and transport failures retain their warnings. Partial
@@ -3543,6 +3557,7 @@ are omitted rather than framing the wrong part of the globe.
 - Build gate: `npm run build`
 - Network access: local-only by default (`HOST=localhost` in dev-fresh.sh); LAN is an explicit opt-in via `HOST=0.0.0.0` (launcher prints a key-exposure warning + LAN URL; see SECURITY.md)
 - OpenSky default mode: OAuth (`OPENSKY_AUTH_MODE=oauth`; `anon` works without credentials)
+- Setup doctor resolves `OPENSKY_AUTH_MODE` from the environment and dotenv files. Explicit `anon` and OAuth mode without a client pair report keyless anonymous access (rate-limited); a complete OAuth pair retains the existing presence-only capability wording. Basic and auto modes report the selected mode without guessing which credentials runtime will accept. The proxy's auth behavior is unchanged.
 - Google key expected in Keychain service `google-maps-api` (or `GOOGLE_MAPS_API_KEY`, or `.env`)
 - OpenSky credentials expected in Keychain service `opensky-network` (or env, or `.env`); `OPENSKY_AUTH_MODE` and `OPENSKY_CREDENTIALS_FILE` read from `.env` too
 - Optional-key precedence in `dev-fresh.sh` is uniform — explicit shell env, then `.env`, then Keychain: `OPENAI_API_KEY` (Keychain `openai-api`/`api-key` — voice + HUD summary), `AISSTREAM_API_KEY` (`aisstream-api`/`api-key` — live vessels), `CESIUM_ION_TOKEN` (`cesium-ion`/`token` — Bing stacks), `TOMTOM_API_KEY` (`tomtom-api`/`api-key` — live traffic flow), `FIRMS_MAP_KEY` (`firms-map`/`map-key` — live fires), `LL2_API_TOKEN` (`.env` only)
@@ -3558,6 +3573,7 @@ are omitted rather than framing the wrong part of the globe.
 - A cold OpenSky failure uses the current camera subpoint only to request a cached adsb.lol point fallback capped at 250 nm. A fresh OpenSky response or last-good cache wins; a nominally successful worldwide snapshot more than two minutes old prefers viewport-scoped adsb.lol when available, otherwise the stale source is reported honestly. The fallback is visibly source-labeled and is never presented as a worldwide snapshot.
 - GBFS proxy refuses upstream redirects (`redirect: 'manual'`; any 3xx becomes a 502 and the redirect target is logged server-side only) and enforces its 5 MB response cap while the body streams, cancelling the upstream read past the cap; CCTV health map is bounded.
 - Proxy error payloads are sanitized (no internal error details returned to clients).
+- That holds for the OpenAI and CCTV media paths too: `/api/openai/hud-summary` never relays OpenAI's own `error.message`, `/api/realtime/token` passes successful ephemeral-token responses through but answers with a fixed error when minting fails or upstream rejects the request, and a failed CCTV media fetch stores a fixed camera health `message` — `GET /api/cctv/health` serializes that field and the CCTV panel renders it as a status label, so it is a client surface as much as the response body is.
 - `OPENAI_API_KEY` is server-side only; the browser receives ephemeral Realtime client secrets from `/api/realtime/token`.
 - `AISSTREAM_API_KEY` is server-side only; the browser reads the same-origin `/api/ais-live` cache.
 - `/api/google/nearby-places` keeps the Google key out of Places requests issued for voice scene context.
@@ -3567,6 +3583,7 @@ are omitted rather than framing the wrong part of the globe.
 - `/api/route` proxies bounded OSRM route requests for annotation routes, with profile allowlisting, distance caps, response caps, caching, and sanitized "no route found" errors.
 - Track endpoints: `/api/ais-live/track?mmsi=` (server-accumulated ring buffers; sub-route handled before the rows snapshot), `/api/opensky-track?icao24=` (OAuth, 60s cache, sanitized errors, independent OpenSky credit bucket), `/api/adsblol/trace?hex=` (60s cache, 5MB cap, ODbL attribution required in UI).
 - Realtime debug logs redact API keys, bearer tokens, client secrets, and image data URLs before writing to disk; request bodies are size-capped.
+- `/api/realtime/debug-log` enforces an always-on 120-per-client/400-global one-minute limiter (not the opt-in `GEV_RATELIMIT_OPENAI_PER_MIN` bucket the cost-bearing OpenAI routes share), appends asynchronously through a serialized queue, and rotates `realtime-conversations.jsonl` at 32 MB keeping one prior generation, so the sink is bounded at twice that regardless of session length. A malformed record answers 400 and a failed write 500, both with a fixed message.
 
 ## UI/UX Runtime Defaults
 
@@ -3800,7 +3817,8 @@ easier to meet (detection is now on more often), but does not create it.
 - Traffic runs in `sim` mode (white dots, hardcoded speeds) unless `TOMTOM_API_KEY`
   is configured (env or Keychain `tomtom-api`/`api-key`), which enables `live` mode:
   TomTom flow vector tiles via the budget-governed `/api/tomtom` proxy
-  (`.gev-cache/tomtom/`, 120 s TTL, `TOMTOM_DAILY_TILE_BUDGET` default 40k/day),
+  (`.gev-cache/tomtom/`, 120 s TTL, `TOMTOM_DAILY_TILE_BUDGET` default 6k/day,
+  sized so a 31-day month stays inside TomTom's 200K/month free allowance),
   decoded client-side (`flowTiles.js`), matched onto Overpass roads
   (`flowMatch.js`), and rendered as green/amber/red dot color + speed/density
   scaling (`trafficFlowStyle.js`); closures spawn no dots; unmatched roads stay
diff --git a/scripts/format-scope.json b/scripts/format-scope.json
index caffd41d2..2ed4f1081 100644
--- a/scripts/format-scope.json
+++ b/scripts/format-scope.json
@@ -495,6 +495,7 @@
   "src/layers/cctv/source.test.mjs",
   "src/data/alprCameras.js",
   "src/data/alprCameras.test.mjs",
+  "src/data/dataCredits.test.mjs",
   "src/layers/alpr/index.js",
   "src/layers/alpr/model.js",
   "src/layers/alpr/policy.js",
@@ -743,5 +744,6 @@
   "src/director/interactions/interactions.test.mjs",
   "scripts/qa-director-interactions.mjs",
   "src/director/sharing/sharing.test.mjs",
-  "scripts/qa-director-sharing.mjs"
+  "scripts/qa-director-sharing.mjs",
+  "src/data/aisStreamSentinels.test.mjs"
 ]
diff --git a/scripts/setup-doctor.mjs b/scripts/setup-doctor.mjs
index 51b78be7d..954369030 100644
--- a/scripts/setup-doctor.mjs
+++ b/scripts/setup-doctor.mjs
@@ -8,6 +8,8 @@ import { projectRoot } from './project-root.mjs';
 import { selectMapStartupRoute } from '../src/mapStartup.js';
 
 const ROOT = projectRoot(import.meta.url);
+const OPENSKY_AUTH_MODE_DEFAULT = 'oauth';
+const OPENSKY_AUTH_MODES = new Set(['basic', 'oauth', 'auto', 'anon']);
 
 export const CREDENTIALS = Object.freeze([
   { name: 'GOOGLE_MAPS_API_KEY', label: 'Google Maps', keychain: [['google-maps-api', 'api-key'], ['google-maps-api', 'default'], ['google-maps-api', 'key']] },
@@ -101,6 +103,21 @@ export function readDoctorDotenvValue(
   return String(values[key] ?? '');
 }
 
+/** Resolve the OpenSky auth mode using the same environment and dotenv precedence as the launcher. */
+export function resolveOpenSkyAuthMode({
+  authoritativeEnvironment = false,
+  environment = process.env,
+  rootDir = ROOT,
+} = {}) {
+  const environmentDefinesMode = Object.prototype.hasOwnProperty.call(environment, 'OPENSKY_AUTH_MODE');
+  const environmentMode = String(environment.OPENSKY_AUTH_MODE || '').trim();
+  const dotenvMode = authoritativeEnvironment && environmentDefinesMode
+    ? ''
+    : readDoctorDotenvValue('OPENSKY_AUTH_MODE', rootDir);
+  const mode = (environmentMode || dotenvMode).trim().toLowerCase();
+  return OPENSKY_AUTH_MODES.has(mode) ? mode : OPENSKY_AUTH_MODE_DEFAULT;
+}
+
 function hasKeychainItem(service, account) {
   if (process.platform !== 'darwin') return false;
   const result = spawnSync('security', [
@@ -128,8 +145,25 @@ export function resolveCredential(spec, {
   return { configured: false, source: null };
 }
 
-export function buildCapabilitySummary(credentials) {
+export function buildCapabilitySummary(
+  credentials,
+  { openSkyAuthMode = OPENSKY_AUTH_MODE_DEFAULT } = {},
+) {
   const configured = (name) => credentials[name]?.configured === true;
+  const hasOAuthCredentials = configured('OPENSKY_CLIENT_ID') && configured('OPENSKY_CLIENT_SECRET');
+  const requestedOpenSkyMode = String(openSkyAuthMode || '').trim().toLowerCase();
+  const openSkyMode = OPENSKY_AUTH_MODES.has(requestedOpenSkyMode)
+    ? requestedOpenSkyMode
+    : OPENSKY_AUTH_MODE_DEFAULT;
+  const flights = openSkyMode === 'anon'
+    ? 'OpenSky keyless anonymous access (rate-limited)'
+    : openSkyMode === 'basic'
+      ? 'OpenSky Basic mode selected (credential presence and validity not verified)'
+      : openSkyMode === 'auto'
+        ? 'OpenSky auto mode selected (runtime credential choice and validity not verified)'
+        : hasOAuthCredentials
+          ? 'OpenSky OAuth credentials present (runtime mode and validity not verified)'
+          : 'OpenSky keyless anonymous access (rate-limited)';
   const route = selectMapStartupRoute({
     googleApiKey: configured('GOOGLE_MAPS_API_KEY') ? 'configured' : '',
     cesiumToken: configured('CESIUM_ION_TOKEN') ? 'configured' : '',
@@ -140,9 +174,7 @@ export function buildCapabilitySummary(credentials) {
       : route === 'google-ion'
         ? 'Google Photorealistic 3D Tiles through Cesium ion; Bing and world-terrain stacks available'
         : 'Esri World Imagery (keyless satellite basemap) with keyless terrain',
-    flights: configured('OPENSKY_CLIENT_ID') && configured('OPENSKY_CLIENT_SECRET')
-      ? 'OpenSky OAuth credentials present (runtime mode and validity not verified)'
-      : 'OpenSky OAuth credentials not configured',
+    flights,
     voice: configured('OPENAI_API_KEY') ? 'available' : 'off until an OpenAI key is added',
     vessels: configured('AISSTREAM_API_KEY') ? 'live AISStream feed' : 'off until an AISStream key is added',
     fires: configured('FIRMS_MAP_KEY') ? 'live NASA FIRMS feed' : 'off until a FIRMS key is added',
@@ -173,7 +205,9 @@ export function inspectSetup({ includeKeychain = true, authoritativeEnvironment
       : { available: false, version: null },
     dependenciesInstalled,
     credentials,
-    capabilities: buildCapabilitySummary(credentials),
+    capabilities: buildCapabilitySummary(credentials, {
+      openSkyAuthMode: resolveOpenSkyAuthMode({ authoritativeEnvironment, rootDir }),
+    }),
   };
 }
 
diff --git a/server/providers/cctv.js b/server/providers/cctv.js
index 5f601532f..85cece004 100644
--- a/server/providers/cctv.js
+++ b/server/providers/cctv.js
@@ -308,11 +308,21 @@ export function cctvProxy({ sourceRoot = process.cwd() } = {}) {
             }
             const timedOut =
               error?.name === 'AbortError' || error?.name === 'TimeoutError';
+            // GET /api/cctv/health serializes `message`, and the CCTV panel
+            // renders it as a status label, so the raw error would leave the
+            // server by a different door than the sanitized body below and
+            // land on screen.
+            //
+            // The log line drops the text too, unlike the catch at the bottom
+            // of this file: a media-fetch failure names the camera's upstream
+            // host, and these lines get pasted into issues. The status codes
+            // below carry the diagnosis — 504 for a timeout, 502 otherwise.
+            console.warn('[CCTV Proxy] media fetch failed');
             setHealth(cameraId, {
               status: 'degraded',
               sourceKind: 'upstream',
               label: source?.provider || 'Configured source',
-              message: error?.message || 'Media fetch failed',
+              message: 'Media fetch failed',
             });
             res.writeHead(timedOut ? 504 : 502, {
               'Content-Type': 'application/json',
diff --git a/server/providers/common/request.js b/server/providers/common/request.js
index ab261d779..93bbbaa85 100644
--- a/server/providers/common/request.js
+++ b/server/providers/common/request.js
@@ -17,17 +17,29 @@ async function readRequestBodyCapped(req, maxBytes) {
 function readRequestBody(req, maxBytes = 1024 * 1024) {
   return new Promise((resolve, reject) => {
     let total = 0;
+    let bodyTooLarge = false;
     const chunks = [];
     req.on('data', (chunk) => {
+      // Keep draining after the cap without retaining more bytes. Destroying
+      // the request also destroys its socket, so the handler cannot deliver
+      // the fixed error response it constructs for BODY_TOO_LARGE.
+      if (bodyTooLarge) return;
       total += chunk.length;
       if (total > maxBytes) {
-        reject(new Error(`Request body exceeds ${maxBytes} bytes`));
-        req.destroy();
+        // Same marker readRequestBodyCapped sets, so a caller can tell an
+        // oversized body from a genuine failure without matching on text.
+        bodyTooLarge = true;
+        chunks.length = 0;
+        const error = new Error(`Request body exceeds ${maxBytes} bytes`);
+        error.code = 'BODY_TOO_LARGE';
+        reject(error);
         return;
       }
       chunks.push(chunk);
     });
-    req.on('end', () => resolve(Buffer.concat(chunks).toString('utf8')));
+    req.on('end', () => {
+      if (!bodyTooLarge) resolve(Buffer.concat(chunks).toString('utf8'));
+    });
     req.on('error', reject);
   });
 }
diff --git a/server/providers/openai/debug-log.js b/server/providers/openai/debug-log.js
index f7dd42154..6d781a31c 100644
--- a/server/providers/openai/debug-log.js
+++ b/server/providers/openai/debug-log.js
@@ -1,16 +1,65 @@
 import { defaultSourceRoot } from '../common/source-root.js';
 import path from 'node:path';
 import { readRequestBody } from '../common/request.js';
-import fs from 'node:fs';
+import { promises as fsp } from 'node:fs';
+import { makeRateLimiter, clientKey } from '../common/rate-limit.js';
 
+/** Cap on one request body. */
 const REALTIME_DEBUG_LOG_MAX_BYTES = 8 * 1024 * 1024;
 
+/**
+ * Ceiling for the live log before it rotates. One generation is kept
+ * (`…jsonl.1`), so the sink is bounded at twice this however long a session
+ * runs. REALTIME_DEBUG_LOG_MAX_BYTES caps a single request; it never bounded
+ * the file those requests accumulate into.
+ */
+const REALTIME_DEBUG_LOG_MAX_FILE_BYTES = 32 * 1024 * 1024;
+
+/**
+ * Per-IP write ceiling. Always on, unlike the opt-in limiter the cost-bearing
+ * OpenAI routes share: throttling those by default would change what a user
+ * spends, while this sink spends disk and event-loop time. A voice session
+ * writes far below this, so the cap needs no configuration to stay invisible.
+ */
+const REALTIME_DEBUG_LOG_MAX_PER_MIN = 120;
+
 function createDebugLogHandler({ sourceRoot = defaultSourceRoot } = {}) {
-  const REALTIME_DEBUG_LOG_DIR = path.join(sourceRoot, '.gev-logs');
-  const REALTIME_DEBUG_LOG_FILE = path.join(
-    REALTIME_DEBUG_LOG_DIR,
-    'realtime-conversations.jsonl',
-  );
+  const logDir = path.join(sourceRoot, '.gev-logs');
+  const logFile = path.join(logDir, 'realtime-conversations.jsonl');
+  const allow = makeRateLimiter({
+    windowMs: 60_000,
+    max: REALTIME_DEBUG_LOG_MAX_PER_MIN,
+    globalMax: 400,
+  });
+
+  // Appends are chained rather than fired concurrently because the rotation
+  // check is a read-then-write on file size: run in parallel, every caller
+  // stats a file still under the ceiling, none of them rotates, and the log
+  // grows past it. The records themselves are O_APPEND and never interleave.
+  let queue = Promise.resolve();
+
+  function append(line) {
+    const done = queue.then(async () => {
+      await fsp.mkdir(logDir, { recursive: true });
+      const size = await fsp.stat(logFile).then(
+        (stat) => stat.size,
+        () => 0,
+      );
+      if (size + Buffer.byteLength(line) > REALTIME_DEBUG_LOG_MAX_FILE_BYTES) {
+        // Node does not replace an existing destination on Windows. Remove the
+        // retained generation explicitly so every rotation behaves the same on
+        // supported platforms rather than failing after the first 64 MB.
+        await fsp.rm(`${logFile}.1`, { force: true });
+        await fsp.rename(logFile, `${logFile}.1`);
+      }
+      await fsp.appendFile(logFile, line);
+    });
+    // The caller still sees its own rejection, but one failed append must not
+    // poison the chain for every request queued behind it.
+    queue = done.catch(() => {});
+    return done;
+  }
+
   return async (req, res) => {
     if (req.method !== 'POST') {
       res.statusCode = 405;
@@ -19,12 +68,19 @@ function createDebugLogHandler({ sourceRoot = defaultSourceRoot } = {}) {
       return;
     }
 
+    if (!allow(clientKey(req))) {
+      res.writeHead(429, {
+        'Content-Type': 'application/json',
+        'Retry-After': '60',
+      });
+      res.end(JSON.stringify({ error: 'Rate limit exceeded' }));
+      return;
+    }
+
     try {
       const body = await readRequestBody(req, REALTIME_DEBUG_LOG_MAX_BYTES);
       const record = JSON.parse(body || '{}');
-      fs.mkdirSync(REALTIME_DEBUG_LOG_DIR, { recursive: true });
-      fs.appendFileSync(
-        REALTIME_DEBUG_LOG_FILE,
+      await append(
         `${JSON.stringify({
           loggedAt: new Date().toISOString(),
           ...record,
@@ -33,13 +89,16 @@ function createDebugLogHandler({ sourceRoot = defaultSourceRoot } = {}) {
       res.statusCode = 204;
       res.end();
     } catch (error) {
-      res.statusCode = 400;
+      // A malformed record or an oversized body is the caller's fault; a failed
+      // write is ours, and only the second is worth a log line. Neither answer
+      // carries the error text, which for an fs failure is an errno and an
+      // absolute path.
+      const callerFault =
+        error instanceof SyntaxError || error?.code === 'BODY_TOO_LARGE';
+      if (!callerFault) console.warn('[realtime-debug-log] append failed');
+      res.statusCode = callerFault ? 400 : 500;
       res.setHeader('Content-Type', 'application/json');
-      res.end(
-        JSON.stringify({
-          error: error?.message || 'Failed to write Realtime debug log',
-        }),
-      );
+      res.end(JSON.stringify({ error: 'Failed to write Realtime debug log' }));
     }
   };
 }
diff --git a/server/providers/openai/hud-summary.js b/server/providers/openai/hud-summary.js
index 54c065412..92d38f2e8 100644
--- a/server/providers/openai/hud-summary.js
+++ b/server/providers/openai/hud-summary.js
@@ -78,20 +78,23 @@ async function handleHudSummary(req, res) {
     res.statusCode = response.ok && summary ? 200 : response.status || 502;
     res.setHeader('Content-Type', 'application/json; charset=utf-8');
     res.setHeader('Cache-Control', 'no-store');
+    if (!response.ok)
+      console.warn(`[hud-summary] upstream HTTP ${response.status}`);
     res.end(
       JSON.stringify({
         summary: summary || null,
-        error: response.ok
-          ? null
-          : data.error?.message || 'OpenAI HUD summary request failed',
+        // Never relay `data.error.message`: that is OpenAI's own wording, and
+        // it carries request ids, organization hints and quota phrasing.
+        error: response.ok ? null : 'OpenAI HUD summary request failed',
       }),
     );
-  } catch (error) {
+  } catch {
+    console.warn('[hud-summary] request failed');
     res.statusCode = 502;
     res.setHeader('Content-Type', 'application/json');
     res.end(
       JSON.stringify({
-        error: error?.message || 'OpenAI HUD summary request failed',
+        error: 'OpenAI HUD summary request failed',
       }),
     );
   }
diff --git a/server/providers/openai/realtime.js b/server/providers/openai/realtime.js
index 53c9d60da..b4dfb4b5a 100644
--- a/server/providers/openai/realtime.js
+++ b/server/providers/openai/realtime.js
@@ -132,26 +132,35 @@ function createRealtimeTokenHandler({
       });
       const body = await response.text();
       res.statusCode = response.status;
-      res.setHeader(
-        'Content-Type',
-        response.headers.get('content-type') || 'application/json',
-      );
       // Which tier/model this secret was actually minted for. The upstream
-      // body is passed through untouched (the client parses it verbatim), so
-      // these headers are the authoritative echo — including the case where a
-      // bogus ?tier= was silently downgraded to standard.
+      // success body is passed through untouched (the client parses it
+      // verbatim), so these headers are the authoritative echo — including the
+      // case where a bogus ?tier= was silently downgraded to standard.
       res.setHeader('X-GEV-Voice-Tier', tier);
       res.setHeader('X-GEV-Voice-Model', model);
       if (requestedTier && !isKnownVoiceTier(requestedTier)) {
         res.setHeader('X-GEV-Voice-Tier-Fallback', '1');
       }
+      if (!response.ok) {
+        console.warn(`[realtime-token] upstream HTTP ${response.status}`);
+        res.setHeader('Content-Type', 'application/json; charset=utf-8');
+        res.end(JSON.stringify({ error: 'Failed to create Realtime token' }));
+        return;
+      }
+      res.setHeader(
+        'Content-Type',
+        response.headers.get('content-type') || 'application/json',
+      );
       res.end(body);
-    } catch (error) {
+    } catch {
+      // For a network fault this was a resolver message naming the upstream
+      // host; the client only needs to know the mint failed.
+      console.warn('[realtime-token] mint failed');
       res.statusCode = 502;
       res.setHeader('Content-Type', 'application/json');
       res.end(
         JSON.stringify({
-          error: error?.message || 'Failed to create Realtime token',
+          error: 'Failed to create Realtime token',
         }),
       );
     }
diff --git a/server/providers/traffic.js b/server/providers/traffic.js
index fda118753..2a3c2e60a 100644
--- a/server/providers/traffic.js
+++ b/server/providers/traffic.js
@@ -23,9 +23,16 @@ import {
  * Budget governor (mirrors the OpenSky credit-governor philosophy — last-good
  * data beats a dead layer): a persistent counter (.gev-cache/tomtom/budget.json,
  * keyed by UTC date, reset on day change) counts upstream fetch attempts
- * against a soft cap (TOMTOM_DAILY_TILE_BUDGET, default 40,000 of the free
- * tier's ~50k/day). Over the cap the proxy serves stale tiles when available,
- * else 429 {error:'budget'}.
+ * against a soft cap (TOMTOM_DAILY_TILE_BUDGET). Over the cap the proxy
+ * serves stale tiles when available, else 429 {error:'budget'}.
+ *
+ * The default is derived from TomTom's published free allowance, which is
+ * granted MONTHLY, not daily: 200,000 Traffic Flow & Incidents tile requests
+ * per month (https://docs.tomtom.com/pricing/). The previous 40,000/day
+ * default — and the "~50k/day" figure this comment used to cite — exhausted
+ * a month's entire allowance in five days, leaving the traffic layer dead
+ * for the rest of the billing period. 6,000/day keeps a full 31-day month
+ * inside the allowance (186,000) with headroom to spare.
  *
  * GET /api/tomtom/status → {hasKey, dailyCount, budget, date}. Keyless mode:
  * status reports hasKey:false and the tile endpoint 503s {error:'no_key'}
@@ -37,7 +44,8 @@ export function tomtomProxy() {
   const TILE_TTL_MS = 120_000;
   const CACHE_DIR = path.join(process.cwd(), '.gev-cache', 'tomtom');
   const BUDGET_PATH = path.join(CACHE_DIR, 'budget.json');
-  const DEFAULT_DAILY_BUDGET = 40000;
+  // 200,000/month free tier / 31 days = 6,451. Rounded down for headroom.
+  const DEFAULT_DAILY_BUDGET = 6000;
   const MEM_MAX_ENTRIES = 256;
   const UPSTREAM_TIMEOUT_MS = 15000;
 
diff --git a/server/providers/vessels/ais-store.js b/server/providers/vessels/ais-store.js
index 4a53464f6..297081b38 100644
--- a/server/providers/vessels/ais-store.js
+++ b/server/providers/vessels/ais-store.js
@@ -72,8 +72,8 @@ export function ingestAisStreamEnvelope(envelope) {
     imo: stringValue(message.ImoNumber ?? message.IMO ?? staticData.imo),
     type: vesselTypeFromAis(message, staticData),
     destination: stringValue(message.Destination ?? staticData.destination),
-    speed: numberValue(message.Sog ?? message.SOG),
-    course: numberValue(message.Cog ?? message.COG),
+    speed: normalizedSpeedOverGround(message.Sog ?? message.SOG),
+    course: normalizedCourseOverGround(message.Cog ?? message.COG),
     heading: normalizedHeading(message.TrueHeading ?? message.Heading),
     last_position_UTC: normalizeAisTimestamp(
       metadata.time_utc ?? metadata.TimeUtc,
@@ -277,6 +277,28 @@ function normalizedHeading(value) {
   return heading !== null && heading >= 0 && heading <= 360 ? heading : null;
 }
 
+/**
+ * Speed over ground in knots, or null when the report carries AIS's "not
+ * available" code. ITU-R M.1371 encodes SOG in 0.1-knot units where 1022 means
+ * "102.2 knots or higher" and 1023 means unavailable, so 102.3 is a sentinel
+ * and not a reading. It is finite, so every consumer would otherwise treat it
+ * as a genuine speed.
+ */
+function normalizedSpeedOverGround(value) {
+  const speed = numberValue(value);
+  return speed !== null && speed >= 0 && speed < 102.3 ? speed : null;
+}
+
+/**
+ * Course over ground in degrees, or null when the report carries AIS's "not
+ * available" code (3600 in 0.1-degree units, i.e. 360). Valid courses stop at
+ * 359.9; heading keeps its own separate sentinel of 511.
+ */
+function normalizedCourseOverGround(value) {
+  const course = numberValue(value);
+  return course !== null && course >= 0 && course < 360 ? course : null;
+}
+
 function normalizeAisTimestamp(value) {
   const text = stringValue(value);
   if (!text) return new Date().toISOString();
diff --git a/src/data/aisStreamSentinels.test.mjs b/src/data/aisStreamSentinels.test.mjs
new file mode 100644
index 000000000..6595b795e
--- /dev/null
+++ b/src/data/aisStreamSentinels.test.mjs
@@ -0,0 +1,65 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import {
+  aisStreamRows,
+  ingestAisStreamEnvelope,
+} from '../../server/providers/vessels/ais-store.js';
+
+// ITU-R M.1371 encodes speed over ground in 0.1-knot units and course over
+// ground in 0.1-degree units, reserving the top code of each field for "not
+// available": SOG 1023 arrives as 102.3 knots and COG 3600 as 360 degrees.
+// Both are finite numbers, so they reach consumers as ordinary readings unless
+// the store rejects them here.
+const positionReport = (mmsi, body) => ({
+  MessageType: 'PositionReport',
+  Message: { PositionReport: { UserID: Number(mmsi), ...body } },
+  MetaData: {
+    MMSI: mmsi,
+    latitude: 53.40879,
+    longitude: 6.19741,
+    time_utc: '2026-09-17 20:00:00.000000000 +0000 UTC',
+  },
+});
+
+const rowFor = (mmsi) => aisStreamRows(50000).find((row) => row.mmsi === mmsi);
+
+test('speed over ground reported as unavailable is unknown, not 102.3 knots', () => {
+  assert.equal(
+    ingestAisStreamEnvelope(positionReport('244060365', { Sog: 102.3 })),
+    true,
+  );
+  assert.equal(rowFor('244060365').speed, null);
+});
+
+test('course over ground reported as unavailable is unknown, not 360 degrees', () => {
+  assert.equal(
+    ingestAisStreamEnvelope(positionReport('245382000', { Cog: 360 })),
+    true,
+  );
+  assert.equal(rowFor('245382000').course, null);
+});
+
+test('genuine speed and course readings survive the unavailable check', () => {
+  ingestAisStreamEnvelope(
+    positionReport('341260000', { Sog: 12.9, Cog: 74.6 }),
+  );
+  const row = rowFor('341260000');
+  assert.equal(row.speed, 12.9);
+  assert.equal(row.course, 74.6);
+});
+
+test('a stopped vessel keeps its zero speed and zero course', () => {
+  ingestAisStreamEnvelope(positionReport('244374235', { Sog: 0, Cog: 0 }));
+  const row = rowFor('244374235');
+  assert.equal(row.speed, 0);
+  assert.equal(row.course, 0);
+});
+
+test('the fastest encodable reading below the unavailable code is kept', () => {
+  ingestAisStreamEnvelope(
+    positionReport('244060366', { Sog: 102.2, Cog: 359.9 }),
+  );
+  const row = rowFor('244060366');
+  assert.equal(row.speed, 102.2);
+  assert.equal(row.course, 359.9);
+});
diff --git a/src/data/dataCredits.js b/src/data/dataCredits.js
index 76029ace8..dde7d3b62 100644
--- a/src/data/dataCredits.js
+++ b/src/data/dataCredits.js
@@ -42,6 +42,17 @@ export const DATA_CREDITS = [
       '<a href="https://adsb.lol" target="_blank" rel="noopener">adsb.lol</a> ' +
       '(ODbL 1.0)',
   },
+  {
+    key: 'adsbdb',
+    html:
+      'Aircraft type, registration &amp; flight routes: ' +
+      '<a href="https://www.adsbdb.com" target="_blank" rel="noopener">adsbdb</a> ' +
+      '· aircraft data from PlaneBase · ICAO-to-N-number conversion by ' +
+      'Guillaume Michel · route data is the work of ' +
+      'David Taylor, Edinburgh, and Jim Mason, Glasgow, and may not be ' +
+      'copied, published, or incorporated into other databases without the ' +
+      'explicit permission of David J Taylor, Edinburgh',
+  },
   {
     key: 'aisstream',
     html:
diff --git a/src/data/dataCredits.test.mjs b/src/data/dataCredits.test.mjs
new file mode 100644
index 000000000..5172867c4
--- /dev/null
+++ b/src/data/dataCredits.test.mjs
@@ -0,0 +1,39 @@
+import { test } from 'node:test';
+import assert from 'node:assert/strict';
+import { DATA_CREDITS } from './dataCredits.js';
+
+test('every credit carries a unique key and some markup to render', () => {
+  const keys = DATA_CREDITS.map((entry) => entry.key);
+  assert.equal(
+    new Set(keys).size,
+    keys.length,
+    'a duplicate key would silently shadow one provider’s credit',
+  );
+  for (const entry of DATA_CREDITS) {
+    assert.ok(entry.key, 'a credit without a key cannot be registered');
+    assert.ok(
+      entry.html && entry.html.trim().length > 0,
+      `credit ${entry.key} has nothing to show`,
+    );
+  }
+});
+
+test('adsbdb is credited and carries its published route-data restriction', () => {
+  const credit = DATA_CREDITS.find((entry) => entry.key === 'adsbdb');
+  assert.ok(
+    credit,
+    'adsbdb supplies aircraft type and routes and must be credited',
+  );
+  // adsbdb publishes this restriction for its route data. Pin the provider's
+  // credits and restriction here so a later edit cannot silently remove them.
+  assert.match(credit.html, /David Taylor, Edinburgh/);
+  assert.match(credit.html, /Jim Mason, Glasgow/);
+  assert.match(
+    credit.html,
+    /may not be\s+copied, published, or incorporated into other databases/,
+  );
+  assert.match(credit.html, /explicit permission of David J Taylor, Edinburgh/);
+  assert.match(credit.html, /PlaneBase/);
+  assert.match(credit.html, /Guillaume Michel/);
+  assert.match(credit.html, /href="https:\/\/www\.adsbdb\.com"/);
+});
diff --git a/src/data/trafficBounds.js b/src/data/trafficBounds.js
index 4bd80ff32..ff5142133 100644
--- a/src/data/trafficBounds.js
+++ b/src/data/trafficBounds.js
@@ -123,12 +123,51 @@ export function deriveFetchCenter({
   return { lat: pulled.lat, lon: pulled.lon, source: 'pulled' };
 }
 
+/**
+ * Wrapped longitude span between two meridians, in degrees.
+ *
+ * Longitude is cyclic, so a box crossing the antimeridian (±180°) has
+ * `east < west` (e.g. `west = 179.98, east = -179.98`). A plain `east - west`
+ * goes negative in that case; wrap through 360° instead.
+ *
+ * @param {number} west - West edge longitude (degrees).
+ * @param {number} east - East edge longitude (degrees).
+ * @returns {number} Span in degrees, always >= 0.
+ */
+function longitudeSpanDeg(west, east) {
+  const raw = east - west;
+  return raw >= 0 ? raw : raw + 360;
+}
+
+/** Keep a monotonic longitude range inside the traffic source's legal domain. */
+function boundedLongitudeRange(center, span) {
+  let west = center - span / 2;
+  let east = center + span / 2;
+  if (east > 180) {
+    west -= east - 180;
+    east = 180;
+  }
+  if (west < -180) {
+    east += -180 - west;
+    west = -180;
+  }
+  return { west, east };
+}
+
 /**
  * Clamp a bounding box's spans to `maxSpanDeg` and recenter it on `center`.
  *
  * Preserves the pre-C4 span semantics (each axis capped at 0.05° ≈ 5.5 km)
  * but centers the box on the derived look-at point instead of the view
- * rectangle's midpoint. Idempotent when `center` is the box's own midpoint.
+ * rectangle's midpoint. At the antimeridian the whole box shifts just far
+ * enough to remain inside [-180, 180], because the road source accepts one
+ * monotonic Overpass bounding box. Idempotent when `center` is the box's own
+ * midpoint.
+ *
+ * Longitude is handled as a cyclic axis for the span calculation, so the
+ * input span wraps correctly across the antimeridian. Output remains
+ * monotonic because callers such as `getBoundsCenter` and `boundsOverlap`
+ * rely on `west <= east`.
  *
  * @param {{south:number, west:number, north:number, east:number}} bounds
  *   Source bounds (span donor).
@@ -138,11 +177,14 @@ export function deriveFetchCenter({
  */
 export function clampBoundsAroundCenter(bounds, center, maxSpanDeg = 0.05) {
   const latSpan = Math.min(bounds.north - bounds.south, maxSpanDeg);
-  const lonSpan = Math.min(bounds.east - bounds.west, maxSpanDeg);
+  const lonSpan = Math.min(
+    longitudeSpanDeg(bounds.west, bounds.east),
+    maxSpanDeg,
+  );
+  const longitude = boundedLongitudeRange(center.lon, lonSpan);
   return {
     south: center.lat - latSpan / 2,
     north: center.lat + latSpan / 2,
-    west: center.lon - lonSpan / 2,
-    east: center.lon + lonSpan / 2,
+    ...longitude,
   };
 }
diff --git a/src/data/trafficBounds.test.mjs b/src/data/trafficBounds.test.mjs
index ec68edba1..0ff38628d 100644
--- a/src/data/trafficBounds.test.mjs
+++ b/src/data/trafficBounds.test.mjs
@@ -108,3 +108,19 @@ test('span clamp is idempotent on already-clamped bounds (loadRoadsForBounds re-
   const twice = clampBoundsAroundCenter(once, midpoint, 0.05);
   assert.deepEqual(twice, once);
 });
+
+test('span clamp keeps antimeridian bounds monotonic and request-safe', () => {
+  for (const centerLon of [179.99, -179.99]) {
+    const result = clampBoundsAroundCenter({
+      south: -0.02,
+      north: 0.02,
+      west: 179.98,
+      east: -179.98,
+    }, { lat: 0, lon: centerLon }, 0.05);
+
+    assert.ok(result.west >= -180 && result.east <= 180);
+    assert.ok(result.west <= result.east, `west/east not monotonic: ${result.west}, ${result.east}`);
+    assert.ok(Math.abs(result.east - result.west - 0.04) < 1e-12);
+    assert.ok(result.west <= centerLon && centerLon <= result.east);
+  }
+});
diff --git a/src/layers/traffic/source.test.mjs b/src/layers/traffic/source.test.mjs
index 763dfe7bf..8047ddbbd 100644
--- a/src/layers/traffic/source.test.mjs
+++ b/src/layers/traffic/source.test.mjs
@@ -2,6 +2,7 @@ import assert from 'node:assert/strict';
 import test from 'node:test';
 import { readFileSync } from 'node:fs';
 import { createTrafficSource } from './source.js';
+import { clampBoundsAroundCenter } from '../../data/trafficBounds.js';
 const bounds = { south: 30.267, west: -97.744, north: 30.268, east: -97.743 };
 const fixture = readFileSync(
   new URL(
@@ -87,6 +88,28 @@ test('road requests have finite bounds and retain the two-pass query', async ()
     /residential/,
   );
 });
+test('antimeridian clamps produce road bounds accepted on either side', async () => {
+  const calls = [];
+  const source = createTrafficSource({
+    fetchImpl: async (...args) => {
+      calls.push(args);
+      return new Response('{"elements":[]}');
+    },
+  });
+  for (const centerLon of [179.99, -179.99]) {
+    const clamped = clampBoundsAroundCenter(
+      {
+        south: -0.02,
+        north: 0.02,
+        west: 179.98,
+        east: -179.98,
+      },
+      { lat: 0, lon: centerLon },
+    );
+    await source.requestRoads(clamped);
+  }
+  assert.equal(calls.length, 2);
+});
 test('malformed availability is an unavailable source rather than a keyless response', async () => {
   const source = createTrafficSource({
     fetchImpl: async () => new Response('{}'),
diff --git a/src/setupDoctor.test.mjs b/src/setupDoctor.test.mjs
index 857a96c78..c66d9e686 100644
--- a/src/setupDoctor.test.mjs
+++ b/src/setupDoctor.test.mjs
@@ -12,6 +12,7 @@ import {
   isConfiguredValue,
   npmProcessSpec,
   readDoctorDotenvValue,
+  resolveOpenSkyAuthMode,
   resolveCredential,
 } from '../scripts/setup-doctor.mjs';
 
@@ -145,6 +146,67 @@ test('doctor reads the dotenv ladder without requiring Vite to be installed', ()
   }
 });
 
+test('doctor resolves OpenSky auth mode with launcher-compatible precedence', () => {
+  const root = mkdtempSync(path.join(tmpdir(), 'gev-doctor-opensky-mode-'));
+  try {
+    writeFileSync(path.join(root, '.env'), 'OPENSKY_AUTH_MODE=anon\n');
+    assert.equal(resolveOpenSkyAuthMode({ environment: {}, rootDir: root }), 'anon');
+    assert.equal(resolveOpenSkyAuthMode({ environment: { OPENSKY_AUTH_MODE: 'oauth' }, rootDir: root }), 'oauth');
+    assert.equal(resolveOpenSkyAuthMode({ environment: { OPENSKY_AUTH_MODE: '' }, rootDir: root }), 'anon');
+    assert.equal(resolveOpenSkyAuthMode({ environment: { OPENSKY_AUTH_MODE: 'invalid' }, rootDir: root }), 'oauth');
+  } finally {
+    rmSync(root, { recursive: true, force: true });
+  }
+});
+
+test('doctor reports keyless anonymous OpenSky access for explicit anon mode', () => {
+  const credentials = {
+    OPENSKY_CLIENT_ID: { configured: true },
+    OPENSKY_CLIENT_SECRET: { configured: true },
+  };
+  assert.equal(
+    buildCapabilitySummary(credentials, { openSkyAuthMode: 'anon' }).flights,
+    'OpenSky keyless anonymous access (rate-limited)',
+  );
+});
+
+test('doctor retains OAuth capability wording for a complete client pair', () => {
+  const credentials = {
+    OPENSKY_CLIENT_ID: { configured: true },
+    OPENSKY_CLIENT_SECRET: { configured: true },
+  };
+  assert.equal(
+    buildCapabilitySummary(credentials, { openSkyAuthMode: 'oauth' }).flights,
+    'OpenSky OAuth credentials present (runtime mode and validity not verified)',
+  );
+});
+
+test('doctor reports Basic mode without inferring runtime auth from OAuth credentials', () => {
+  const oauthCredentials = {
+    OPENSKY_CLIENT_ID: { configured: true },
+    OPENSKY_CLIENT_SECRET: { configured: true },
+  };
+  for (const credentials of [{}, oauthCredentials]) {
+    assert.equal(
+      buildCapabilitySummary(credentials, { openSkyAuthMode: 'basic' }).flights,
+      'OpenSky Basic mode selected (credential presence and validity not verified)',
+    );
+  }
+});
+
+test('doctor reports auto mode without assuming its eventual credential choice', () => {
+  const oauthCredentials = {
+    OPENSKY_CLIENT_ID: { configured: true },
+    OPENSKY_CLIENT_SECRET: { configured: true },
+  };
+  for (const credentials of [{}, oauthCredentials]) {
+    assert.equal(
+      buildCapabilitySummary(credentials, { openSkyAuthMode: 'auto' }).flights,
+      'OpenSky auto mode selected (runtime credential choice and validity not verified)',
+    );
+  }
+});
+
 test('doctor describes the credential ladder without exposing values', () => {
   const credentials = {
     GOOGLE_MAPS_API_KEY: { configured: false },
@@ -163,7 +225,7 @@ test('doctor describes the credential ladder without exposing values', () => {
   assert.match(capabilities.map, /Bing and world-terrain stacks/);
   assert.equal(capabilities.voice, 'available');
   assert.match(capabilities.missions, /token allowance/);
-  assert.equal(capabilities.flights, 'OpenSky OAuth credentials not configured');
+  assert.equal(capabilities.flights, 'OpenSky keyless anonymous access (rate-limited)');
 
   const report = formatSetupReport({
     ready: true,
diff --git a/src/tooling/localServices.test.mjs b/src/tooling/localServices.test.mjs
index d7f2e4ba3..26eebfa08 100644
--- a/src/tooling/localServices.test.mjs
+++ b/src/tooling/localServices.test.mjs
@@ -1,5 +1,6 @@
 import assert from 'node:assert/strict';
 import test from 'node:test';
+import { createServer } from 'node:http';
 import { Readable } from 'node:stream';
 import {
   mkdtempSync,
@@ -279,7 +280,157 @@ test('Realtime service configuration selects compatible endpoint/model without f
     url: '/?tier=arbitrary-model&model=other',
   });
   assert.equal(response.status, 200);
+  assert.deepEqual(response.json(), { value: 'short-lived-fixture' });
   assert.equal(response.headers['x-gev-voice-model'], 'configured-model');
   assert.equal(response.headers['cache-control'], 'no-store');
   assert.doesNotMatch(response.body, /server-fixture|voice\.example/);
 });
+
+test('OpenAI routes answer generically when the upstream or the request fails', async (t) => {
+  env(t, 'OPENAI_API_KEY', 'fixture-upstream-secret');
+  env(t, 'GEV_RATELIMIT_OPENAI_PER_MIN', undefined);
+  const leak =
+    'fixture-upstream-secret req_fixture_1234 org-fixture quota exhausted';
+
+  // `data.error.message` is OpenAI's own wording — request ids, organization
+  // hints, quota phrasing — and was relayed verbatim whenever upstream was not ok.
+  t.mock.method(globalThis, 'fetch', async () =>
+    Response.json({ error: { message: leak } }, { status: 429 }),
+  );
+  const summary = await request(
+    install(openAiRealtimeProxy()).get('/api/openai/hud-summary'),
+    { method: 'POST', body: JSON.stringify({ context: {} }) },
+  );
+  assert.equal(summary.json().error, 'OpenAI HUD summary request failed');
+  assert.equal(summary.body.includes('req_fixture_1234'), false);
+  assert.equal(summary.body.includes('fixture-upstream-secret'), false);
+
+  // An HTTP error from the client-secret endpoint carries the same upstream
+  // detail, while a successful response must still pass the ephemeral secret.
+  t.mock.restoreAll();
+  t.mock.method(globalThis, 'fetch', async () =>
+    Response.json({ error: { message: leak } }, { status: 429 }),
+  );
+  const rejectedToken = await request(
+    install(openAiRealtimeProxy()).get('/api/realtime/token'),
+  );
+  assert.equal(rejectedToken.status, 429);
+  assert.equal(
+    rejectedToken.headers['content-type'],
+    'application/json; charset=utf-8',
+  );
+  assert.deepEqual(rejectedToken.json(), {
+    error: 'Failed to create Realtime token',
+  });
+  assert.equal(rejectedToken.body.includes('req_fixture_1234'), false);
+  assert.equal(rejectedToken.body.includes('fixture-upstream-secret'), false);
+
+  // A network fault surfaced a resolver message naming the upstream host.
+  t.mock.restoreAll();
+  t.mock.method(globalThis, 'fetch', async () => {
+    throw Error(`getaddrinfo ENOTFOUND api.openai.com ${leak}`);
+  });
+  const token = await request(
+    install(openAiRealtimeProxy()).get('/api/realtime/token'),
+  );
+  assert.equal(token.status, 502);
+  assert.deepEqual(token.json(), { error: 'Failed to create Realtime token' });
+  assert.equal(token.body.includes('api.openai.com'), false);
+  assert.equal(token.body.includes('fixture-upstream-secret'), false);
+});
+
+test('the debug-log sink stays bounded, rate limited, and quiet about failures', async (t) => {
+  const sourceRoot = root(t);
+  const handler = install(openAiRealtimeProxy({ sourceRoot })).get(
+    '/api/realtime/debug-log',
+  );
+  const file = path.join(sourceRoot, '.gev-logs/realtime-conversations.jsonl');
+  const write = (record) =>
+    request(handler, { method: 'POST', body: JSON.stringify(record) });
+
+  // A malformed record is the caller's fault and a 400; neither answer carries
+  // the error text, which for a write failure is an errno and an absolute path.
+  const malformed = await request(handler, { method: 'POST', body: '{nope' });
+  assert.equal(malformed.status, 400);
+  assert.deepEqual(malformed.json(), {
+    error: 'Failed to write Realtime debug log',
+  });
+
+  // The limiter is always on, unlike the opt-in one the cost-bearing routes
+  // share: 120/min per IP, far above what a voice session writes.
+  let limited = null;
+  let accepted = 0;
+  for (let n = 0; n < 130 && !limited; n += 1) {
+    const response = await write({ n });
+    if (response.status === 429) limited = response;
+    else if (response.status === 204) accepted += 1;
+  }
+  assert.ok(limited, 'the sink refuses a caller past its per-minute ceiling');
+  assert.equal(limited.headers['retry-after'], '60');
+  assert.deepEqual(limited.json(), { error: 'Rate limit exceeded' });
+  // The limiter counts requests rather than successful writes, so the malformed
+  // record above already spent one of the 120 slots.
+  assert.equal(accepted, 119);
+
+  // Every accepted record is on disk and parses: the queue serializes appends,
+  // so none was lost or truncated by the ones beside it.
+  const lines = readFileSync(file, 'utf8').split('\n').filter(Boolean);
+  assert.equal(lines.length, accepted);
+  for (const line of lines) assert.doesNotThrow(() => JSON.parse(line));
+  assert.ok(lines.every((line) => JSON.parse(line).loggedAt));
+});
+
+test('an oversized debug-log request receives the fixed error response', async (t) => {
+  const handler = install(openAiRealtimeProxy({ sourceRoot: root(t) })).get(
+    '/api/realtime/debug-log',
+  );
+  const server = createServer(handler);
+  await new Promise((resolve) => server.listen(0, '127.0.0.1', resolve));
+  t.after(() => new Promise((resolve) => server.close(resolve)));
+
+  const response = await fetch(
+    `http://127.0.0.1:${server.address().port}/api/realtime/debug-log`,
+    { method: 'POST', body: 'x'.repeat(8 * 1024 * 1024 + 1) },
+  );
+  assert.equal(response.status, 400);
+  assert.deepEqual(await response.json(), {
+    error: 'Failed to write Realtime debug log',
+  });
+});
+
+test('the debug log rotates instead of growing without bound', async (t) => {
+  const sourceRoot = root(t);
+  const handler = install(openAiRealtimeProxy({ sourceRoot })).get(
+    '/api/realtime/debug-log',
+  );
+  const file = path.join(sourceRoot, '.gev-logs/realtime-conversations.jsonl');
+
+  // 8 MB bounds one request body; nothing bounded the file until now, so a
+  // single page could grow it for as long as the dev server ran. Each record
+  // here is ~1 MB, well inside the body cap.
+  const pad = 'x'.repeat(1024 * 1024);
+  // Seventy-two records force two rotations. The second one replaces an
+  // existing `.1`, which requires an explicit removal on Windows.
+  for (let n = 0; n < 72; n += 1) {
+    assert.equal(
+      (
+        await request(handler, {
+          method: 'POST',
+          body: JSON.stringify({ n, pad }),
+        })
+      ).status,
+      204,
+    );
+  }
+
+  const live = statSync(file).size;
+  const previous = statSync(`${file}.1`).size;
+  assert.ok(live <= 32 * 1024 * 1024, `live log under the ceiling (${live})`);
+  assert.ok(
+    live + previous <= 64 * 1024 * 1024,
+    'both generations together stay within twice the ceiling',
+  );
+  // Unrotated, these records would be ~75 MB in one file.
+  assert.ok(live + previous < 64 * 1024 * 1024);
+  assert.ok(!existsSync(`${file}.2`), 'exactly one generation is retained');
+});
diff --git a/src/tooling/mediaProviders.test.mjs b/src/tooling/mediaProviders.test.mjs
index cbfcd0624..9642f55b6 100644
--- a/src/tooling/mediaProviders.test.mjs
+++ b/src/tooling/mediaProviders.test.mjs
@@ -149,3 +149,45 @@ for (const hook of ['configureServer', 'configurePreviewServer']) {
     assert.equal(cancelled, true);
   });
 }
+
+test('a failed CCTV media fetch reports a fixed health message, not the error text', async (t) => {
+  isolate(t);
+  const root = mkdtempSync(path.join(tmpdir(), 'gev-cctv-'));
+  t.after(() => rmSync(root, { recursive: true, force: true }));
+  mkdirSync(path.join(root, 'config'));
+  writeFileSync(
+    path.join(root, 'config/cctv_sources.austin.json'),
+    JSON.stringify([
+      {
+        id: 'leaky',
+        name: 'Leaky camera',
+        lat: 30.27,
+        lon: -97.74,
+        feedType: 'video',
+        url: 'https://cams.fixture.invalid/leaky.m3u8',
+      },
+    ]),
+  );
+  const leak =
+    'connect ETIMEDOUT 203.0.113.9:443 C:\fixture\secret-path\cams.json';
+  t.mock.restoreAll();
+  t.mock.method(globalThis, 'fetch', () => {
+    throw Error(leak);
+  });
+
+  const call = install(cctvProxy({ sourceRoot: root }));
+  const media = await call('/media/leaky');
+  assert.equal(media.status, 502);
+  assert.equal(media.body.includes('203.0.113.9'), false);
+
+  // GET /health is the second door: src/layers/cctv/frames.js renders each
+  // entry's `message` as the camera's status label, so a raw errno stored here
+  // reaches the screen even though the response above is sanitized.
+  const camera = JSON.parse((await call('/health')).body).cameras.find(
+    (entry) => entry.id === 'leaky',
+  );
+  assert.equal(camera.status, 'degraded');
+  assert.equal(camera.message, 'Media fetch failed');
+  assert.equal(camera.message.includes('ETIMEDOUT'), false);
+  assert.equal(camera.message.includes('secret-path'), false);
+});
diff --git a/src/ui/styles/foundation.css b/src/ui/styles/foundation.css
index 828eb9922..2d374b2c6 100644
--- a/src/ui/styles/foundation.css
+++ b/src/ui/styles/foundation.css
@@ -14,6 +14,8 @@
   --text-primary: #e8eaed;
   --text-secondary: rgba(232, 234, 237, 0.5);
   --text-dim: rgba(232, 234, 237, 0.3);
+  /* Opaque surface for browser-painted menus, which cannot use --glass-bg. */
+  --menu-bg: #12121c;
   --font-mono: 'JetBrains Mono', 'SF Mono', 'Fira Code', monospace;
   --font-sans: 'Inter', -apple-system, BlinkMacSystemFont, sans-serif;
   --panel-radius: 16px;
@@ -69,6 +71,26 @@
   outline-offset: 2px;
 }
 
+/* Native `<select>` popups are painted by the browser, not by this sheet: the
+   closed control inherits its panel skin, but the option list falls back to the
+   platform's light palette unless the control declares a scheme. Declare it on
+   the controls rather than on :root so the app keeps its own hand-tuned
+   scrollbar and input skins, and set the option colors explicitly because the
+   popup surface is not themed on every platform. */
+select {
+  color-scheme: dark;
+}
+
+select option,
+select optgroup {
+  background-color: var(--menu-bg);
+  color: var(--text-primary);
+}
+
+select option:disabled {
+  color: var(--text-dim);
+}
+
 html,
 body {
   width: 100%;