docs(performance): cables rewrite in #1722 and clouds shader in #1721

Co-Authored-By: Claude Fable 5.2 <[email protected]>

DATA_SOURCES.md

@@ -1,13 +1,13 @@
 # Data Sources & Attribution
 
-God's Eye View's **code** is [MIT](LICENSE)-licensed. **The MIT grant covers the source code only — it does NOT extend to third-party data or visual assets.** Every third-party source keeps its own license and terms. This file documents the live and bundled data sources. The bundled 3D models are the project's own procedural fleet models, MIT-licensed with the code ([`public/models/README.md`](public/models/README.md)).
+God's Eye View's **code** is [MIT](LICENSE)-licensed. **The MIT grant covers the source code only — it does NOT extend to third-party data or visual assets.** Every third-party source keeps its own license and terms. This file documents the live and bundled data sources; bundled 3D-model provenance is recorded in [`public/models/README.md`](public/models/README.md).
 
 How to read this:
 
 - **The non-permissive datasets are carved out, not omitted.** Some bundled data (e.g. TeleGeography, CC BY-NC-SA) isn't MIT-compatible. Rather than hide it, we **bundle it with a clear license carve-out** so the app works out of the box — but it stays under the provider's terms.
 - **If your use doesn't fit a dataset's license, remove that dataset.** Most importantly: TeleGeography is **NonCommercial** — commercial users must delete it (or license it from TeleGeography). It's one self-contained folder.
 - **Attribution is shown in-app** and listed here. Keep it intact. The required Google/Cesium credit renders on the on-globe credit line (bottom-left, `#cesium-credits`), and every per-layer credit below is registered into the expandable **"Data attribution"** lightbox on that line (`src/data/dataCredits.js` → `viewer.creditDisplay.addStaticCredit`). Both stay visible in clean-view and recording modes.
-- **The bundled 3D models are the project's own.** Every model under `public/models/` is a procedural fleet model generated by the project maintainers and covered by the repository's MIT license; [`public/models/README.md`](public/models/README.md) describes the set.
+- **Bundled model attribution lives beside the model files.** [`public/models/README.md`](public/models/README.md) records each shipped model's creator, source, license, and modification status.
 
 ---
 
@@ -564,11 +564,8 @@ Global Forecast System (GFS). The `/api/wind` server-side proxy selects the
 latest available 0.25° cycle, reads its `.idx` inventory, byte-range fetches only
 the `UGRD`/`VGRD` 10 m GRIB2 messages, optionally adding `TMP` at 2 m or
 `PRMSL` at mean sea level from the same run and forecast time. It decodes the
-selected messages with ecCodes (WASM) and resamples them to a compact grid (1° by
-default) on a worker thread, so a refresh does not stall other API routes, and
-serves a manifest plus Float32 U/V and optional scalar values. The sampled wind
-of the current forecast file is kept, so choosing a color field fetches only
-that field. It is keyless, cached
+selected messages with ecCodes (WASM), resamples to a compact grid (1° by
+default), and serves a manifest plus Float32 U/V and optional scalar values. It is keyless, cached
 per cycle for an hour, and is a **forecast, not an observation**: the particles
 show model flow, not measured wind. The renderer bakes bounded curves through
 the sampled field and animates their phase on the GPU, with a canvas fallback.

LICENSE

@@ -56,6 +56,7 @@ In short: the MIT grant does not extend to any third-party data. See
 DATA_SOURCES.md for the full per-source license + attribution summary, and
 remove any dataset whose terms don't fit your use.
 
-The 3D models under public/models/ are the project's procedural fleet
-models, generated from code by the project maintainers, and are covered by
-the MIT license above like the source code. See public/models/README.md.
+The 3D models under public/models/ are also third-party works and are NOT
+licensed under MIT. They remain under their individual licenses. See
+public/models/README.md for the author, source, license, modification notice,
+and attribution covering each model file.

README.md

@@ -60,7 +60,7 @@ Start with the included data sources, then add your own. Each layer is a separat
 - **📡 Contacts:** A 250 km roster of everything near your target — step through live aircraft and drop into any cockpit.
 - **🎯 Click-to-track anything:** Camera locks on, draws a fading trail, surfaces full metadata — and a tracked fire or vessel hands you off to the nearest live camera in one click.
 - **🖊️ Voice whiteboard:** Speak annotations onto the world — real boundary polygons, marks, and routes.
-- **🛫 3D fleet:** 121 procedural models drawn life size by type — airliners, business jets, helicopters, drones, ships, buses, trams and trains — and every contact swaps from glyph to 3D model as you close in.
+- **🛫 3D hangar:** Real per-class aircraft models — 787, ATR-72, Citation, Bell 206, MQ-9 — and a tracked contact swaps from glyph to 3D model as you close in.
 - **🎨 Reskin reality:** GLSL sensor looks over the normal globe — CRT, NVG, FLIR/thermal, Noir.
 - **🟩 Detection overlay:** Screen-space bounding boxes and IDs on everything in view.
 - **🎖️ Military HUD:** Tactical heads-up display with intelligence-style telemetry.

changelog.d/1698-new-strings-translations.md

@@ -1,5 +0,0 @@
-- Fix the translations of the strings added since the last review (News,
-  Data Library, Wildlife Sightings, Shipwrecks, Rail, the vessel and orbital
-  cockpits) in all twelve languages: neutral side labels and counts no longer
-  break the sentence, terms follow each glossary, and Traditional Chinese
-  follows Simplified Chinese again (German, @gekh, #1698).

changelog.d/1714-fleet-models.md

@@ -1,8 +0,0 @@
-- Draw every 3D model from the project's procedural fleet set: 121 models
-  (about 2 MB, MIT like the code) chosen per vehicle by type and drawn life
-  size, replacing the eight third-party aircraft models. Flights, Military
-  Flights and Local ADS-B pick a model by ICAO type, class and emitter
-  category; Vessels, Transit and Rail now draw models too, under the same 3D
-  toggle and distances, a ship at its AIS length and beam and a tram, bus or
-  train in its mode or category colour. The whole set loads once when a layer
-  starts (samehkhamis, #1714).

changelog.d/1717-wind-server-decode.md

@@ -1,4 +0,0 @@
-- The Wind server no longer blocks every other request while it decodes a
-  forecast. Decoding runs on a worker thread (up to 400 ms of blocking down to
-  about 4 ms), and choosing an overlay after wind has loaded fetches only the
-  overlay's field (mazy1998, #1717).

docs/CURRENT-STATE.md

@@ -1706,7 +1706,7 @@ Applications supply the existing floor/snap, geoid, picking, sprite, camera,
 trail, focus, readout, context, aircraft presentation and render services.
 The factory never constructs a second application registry. Configure a source
 before initialization; replacing it while initialized is rejected. Model loads
-use the supplied asset resolver, including the up-front fleet set load. Enrichment receives
+use the supplied asset resolver, including the preload path. Enrichment receives
 an abort signal and cannot update a later lifecycle after destruction. Existing
 camera, terrain floor, trail, selection and measured model-size policies remain.
 
@@ -2420,9 +2420,8 @@ is validated against Cesium 1.138.0 and must be revalidated on upgrades.
 
 `scripts/qa-anime-sky.mjs` checks asset readiness, visible Normal sky and
 Anime band coverage against a Vite dev server.
-`scripts/qa-sky-model-lighting.mjs` compares a lit fleet airliner
-(`public/models/fleet/aircraft/a320.glb`) against black and white environment
-maps in isolated scenes.
+`scripts/qa-sky-model-lighting.mjs` compares a lit `airplane.glb` against
+black and white environment maps in isolated scenes.
 
 ## Model atmosphere on Apple Metal
 
@@ -3607,22 +3606,20 @@ This is the current runtime/source-of-truth snapshot for the project.
 >   24.09 m aft (70 % of the model's rendered envelope, so inside it), 2.22 m
 >   below centre (6 %), and 2e-9 m off the centreline.
 > - **The tracked trail attaches to REAL HULL, aft and below (2026-08-23):**
->   `FLEET_TRAIL_ANCHORS` (`src/data/fleetTrailAnchors.js`, generated by
->   `scripts/measure-fleet-trail-anchors.mjs`) holds, per fleet aircraft and
->   helicopter model and in RAW glTF coordinates (see the transform-chain entry
->   below), the point of the hull's CENTRELINE PROFILE (its centreline slice)
->   closest to the aft-belly AABB corner. It is NOT the corner: a bounding-box
->   corner is empty space (4.80 m off the nearest triangle on the airliner
->   asset this was first measured on, 6.14 m on the jet), so the trail ended in
->   mid air beside the aircraft. The head segment
+>   `MODEL_TRAIL_ANCHOR_NATIVE` (`src/data/modelVisualAnchor.js`) holds, per GLB
+>   and in RAW glTF coordinates (see the transform-chain entry below), the point
+>   of the hull's CENTRELINE PROFILE (its y = 0 slice) closest to the aft-belly
+>   AABB corner. It is NOT the corner: a bounding-box corner is empty
+>   space, 4.80 m off the nearest triangle on `airplane.glb` and 6.14 m on
+>   `jet.glb`, so the trail ended in mid air beside the aircraft. The head segment
 >   is drawn from a point behind the aircraft to this anchor, so the anchor must
 >   stay aft (83–96 % of each aft extreme) or the segment enters at the tail and
->   stops inside the fuselage. `fleetModels.test.mjs` reads the POSITION BUFFERS —
+>   stops inside the fuselage. `modelScale.test.mjs` reads the POSITION BUFFERS —
 >   real vertices and triangles, because accessor min/max cannot tell a corner
 >   from a surface — and re-derives on-hull, aft, lowest-at-its-station,
 >   on-centreline, and the construction itself. Re-measure, never re-guess.
 > - **The trail anchor rides CESIUM'S transform chain, not a hand-rolled one
->   (2026-08-23 regression fix):** `FLEET_TRAIL_ANCHORS` stores RAW
+>   (2026-08-23 regression fix):** `MODEL_TRAIL_ANCHOR_NATIVE` stores RAW
 >   glTF coordinates and `modelAnchorWorld()` assembles
 >   `modelMatrix × components.transform × axisCorrection` — the same chain
 >   `ModelSceneGraph` renders with, built from Cesium's own exported
@@ -3657,7 +3654,7 @@ This is the current runtime/source-of-truth snapshot for the project.
 >   ENVELOPE, and between them the drawn start slides along the segment so the
 >   visible length grows CONTINUOUSLY from zero. Three earlier cuts are recorded
 >   because their shape matters: testing segment LENGTH against the radius HID
->   real trail (58 m aft of an airliner's 34.41 m envelope gives a 33.98 m
+>   real trail (58 m aft of `airplane.glb`'s 34.41 m envelope gives a 33.98 m
 >   segment, suppressed, though ~23.6 m of it is open air); CLIPPING at the
 >   envelope and keeping only what lay outside stopped a moving trail visibly
 >   SHORT of the aircraft, since a bounding sphere encloses a lot of empty space
@@ -4833,9 +4830,8 @@ activeCameraCardEnabled: true })` may publish it through the retained
 > - **Flights 3D/cockpit hardening:** the destination-direction cue is an
 >   inline SVG rather than a remotely loaded Material Icons ligature, so a
 >   blocked or late font can no longer expand the literal word `navigation`
->   across the cockpit. A grounded aircraft stands its fleet model's scaled
->   clearance (the catalog's `clearance`, measured below the model's origin)
->   above the ground.
+>   across the cockpit. The grounded `airplane.glb` belly offset is calibrated
+>   to the current asset's measured Y-up bounds (`0.063` native units).
 >   Contact Previous/Next and Live Signals aircraft handoffs now
 >   re-seed the first-person camera anchor at the selected flight immediately;
 >   bounded feed-correction smoothing remains scoped to the same aircraft.
@@ -7294,9 +7290,8 @@ pins and share links already supply complete boxes.
   normal four-samples-per-250-ms budget. Retry timestamps more than the maximum
   30-second backoff ahead of the current clock are ignored. A smaller rollback
   can retain a bounded retry delay of up to 30 seconds.
-- Commercial and military aircraft use the same high-level FLEET model regime: 2D billboards when zoomed out, optional glTF models when closer, controlled by the DISPLAY rail `3D` toggle and `Proximity` / `All` modes. Every model is one of the project's procedural fleet models (`public/models/fleet/`, MIT, 121 GLBs, about 2 MB in all; `public/models/README.md`), chosen per aircraft by `aircraftModelSpec` (`src/data/fleetModels.js`): its ICAO type's entry in `types.json`, else the model of its `classifyAircraft` class for a type the map lacks, else its ADS-B emitter category's, else the default airliner. It is drawn LIFE SIZE: the type's published length over the model's built length (`src/layers/flights/modelSpec.js`, shared by Flights, Military Flights and Local ADS-B), and a grounded aircraft stands the model's scaled clearance above the ground. `classifyAircraft` still picks the 2D icons. The whole set loads once, up front, when a layer that draws it starts (`loadFleetModelSet`, which logs `[Fleet] N/121 models (X MB) loaded in N ms`); models are built from those bytes, never fetched mid-session, and an aircraft stays its billboard until the set has loaded. Textured civilians carry a HEAVY tint, not a light one: `MODEL_COLOR_BLEND_AMOUNT` is `0.94` in BOTH layers under Cesium's `ColorBlendMode.MIX`, so the class colour supplies 94% of the surface and the model's own vertex colour ~6%; military models are flat amber, or emergency red. IR boost raises the blend to a full `1.0`. Under NVG/FLIR (map preset or Cockpit vision) models render unlit flat-white at full alpha and scene fog is disabled (fog otherwise blacks out distant models with the globe hidden); state restores on exit.
-- Vessels, Transit and Rail draw fleet models under the same DISPLAY rail `3D` toggle, at the same altitude ceiling, caps and add/keep radii as the aircraft (`FLEET_MODEL_POLICY`, `src/layers/flights/policy.js`; `createFleetModelLayer`, `src/data/fleetModelInstances.js`). A vessel's model follows its AIS ship type and is drawn at its reported length and beam (AIS static dimensions A+B and C+D, now carried on vessel rows as `length` and `beam`), standing at the sea surface, heading by true heading then course. A transit vehicle's model follows its layer mode (bus, tram, subway, rail or ferry; a vehicle of `unknown` mode keeps its icon), a train's its rail category and mode (rail, metro or tram). Neither the transit nor the rail sources carry a GTFS `route_type` or a finer mode (monorail, gondola, funicular, cable car), so the set's models for those and their `transit.elevation` lifts are not reached by any live source today. Both stand on the road or rail top they are drawn on, without the 1.5 m `GROUND_FLOOR_LIFT_M` their icons carry, and are painted in their mode or category colour where the model's `_TINT` attribute marks it. A vessel model is clipped at its waterline (one clipping plane per model, at the model origin's horizontal plane), so its hull below the water never shows through a plain globe, which Cesium does not depth-test models against. Under the NVG/FLIR styles (the Flights `irBoost` parameter) these models render unlit flat white, as aircraft models do. Each vehicle's icon stays the visual until its model is loaded, ready and placed; a model is picked as its icon is, and a vehicle drawn as its model keeps its selection card and its place in the detection overlay.
-- **Aircraft models skip Cesium's dynamic environment map (2026-10-24).** Flights, Military Flights and Local ADS-B pass `environmentMapOptions: AIRCRAFT_ENVIRONMENT_MAP_OPTIONS` (`{ enabled: false }`, `src/data/aircraftModelLighting.js`) to every drawn model. Otherwise each `Model` renders the sky around it and reads it back from the GPU (`readPixels`) when it is first placed, and again whenever it moves more than 1 km, the scene's atmosphere settings change, or (under sunlight dynamic lighting) an hour of scene time passes (`DynamicEnvironmentMapManager`, Cesium 1.138). With the class colour supplying 94% of the surface, the models are lit from Cesium's default spherical harmonics instead, with no specular environment map, so the sky's model-lighting cube map no longer reaches aircraft. `src/data/aircraftModelLighting.test.mjs` pins the option at every drawn model call; every model is built from the fleet set loaded up front, so there is no separate preload.
+- Commercial and military aircraft use the same high-level FLEET model regime: 2D billboards when zoomed out, optional glTF models when closer, controlled by the DISPLAY rail `3D` toggle and `Proximity` / `All` modes. Since 2026-08-16 (Hangar fleet) models are PER-CLASS: real CC-BY GLBs for light/bizjet/turboprop/widebody/helicopter/uav (`CLASS_MODEL_REAL` in `src/data/aircraftClass.js` — meters-baked, scale 1, per-model belly/radius; provenance in `public/models/README.md`), the shared `airplane.glb` for the remaining civilian classes, and the military layer maps weight classes (real GLBs / 747 heavies / `jet.glb` fastjets, per-model heading offsets, flat amber or emergency red). Textured civilians carry a HEAVY tint, not a light one: `MODEL_COLOR_BLEND_AMOUNT` is `0.94` in BOTH layers under Cesium's `ColorBlendMode.MIX`, so the class colour supplies 94% of the surface and the asset's own texture ~6%. The visual direction is clean light silhouettes with only a weak diffuse contribution from the approved textures, so liveries deliberately do NOT read. IR boost raises the blend to a full `1.0`. Under NVG/FLIR (map preset or Cockpit vision) models render unlit flat-white at full alpha and scene fog is disabled (fog otherwise blacks out distant models with the globe hidden); state restores on exit.
+- **Aircraft models skip Cesium's dynamic environment map (2026-10-24).** Flights, Military Flights and Local ADS-B pass `environmentMapOptions: AIRCRAFT_ENVIRONMENT_MAP_OPTIONS` (`{ enabled: false }`, `src/data/aircraftModelLighting.js`) to every drawn model. Otherwise each `Model` renders the sky around it and reads it back from the GPU (`readPixels`) when it is first placed, and again whenever it moves more than 1 km, the scene's atmosphere settings change, or (under sunlight dynamic lighting) an hour of scene time passes (`DynamicEnvironmentMapManager`, Cesium 1.138). With the class colour supplying 94% of the surface, the models are lit from Cesium's default spherical harmonics instead, with no specular environment map, so the sky's model-lighting cube map no longer reaches aircraft. `src/data/aircraftModelLighting.test.mjs` pins the option at every drawn model call; the two preload calls only warm the glTF cache and are never drawn.
 - The DISPLAY-rail `3D` toggle is the user-facing activation path for both aircraft layers. Their small approved GLBs build their render resources without Cesium's frame-spread job queue, preventing continuous Photorealistic 3D Tiles streaming from starving model readiness; model caps, tracking, camera, and fallback billboards are unchanged.
 - **The `3D` toggle DEFAULTS ON in `proximity` on a first run.** Proximity is itself the budget — models appear only below `MODEL_ALT_CEIL_M` and only for the nearest `MODEL_MAX` in view — so the default costs nothing at globe scale, and `all` remains a deliberate opt-in. A fresh boot runs NO layer-state restoration (`LayerStateCoordinator.start()` returns early with neither a share payload nor stored state), so four independent initializers decide what a first-run operator sees and must agree: `booleanOption('models3d', 'e', true)` in `src/data/layerState.js`, `_models3dEnabled = true` in BOTH flight layers, `this._models3dEnabled = true` in `src/ui/aircraftDisplay.js`, and the `active` / `visible` classes on `#models3d-toggle` / `#models3d-mode-row` in `index.html`. All four are pinned together in `src/data/layerState.test.mjs`. Explicit state still wins: because the codec omits default-valued options, `models3d: false` is now what travels in a link (`lo=…f.e.0`) and restores OFF at both aircraft layers. **Consequence for returning users:** a stored `gev:layer-state:v2` blob is a FULL options snapshot, so a session that wrote one before this change carries `models3d:false` and keeps 3D off until the operator flips it (or clears the key) — the durable snapshot is treated as the recipient's own state, by design.
 - **Consequence of the flip on the recorded tracked/fleet inversion:** the 150–800 km band where surrounding contacts draw as models while the SELECTED one is still a glyph is now what an operator sees WITHOUT arming anything. The inversion itself is unchanged and still deliberate (see `src/data/trackedModelRegime.js`); only its reachability changed.

docs/PERFORMANCE-AUDIT-2026-11.md

@@ -0,0 +1,322 @@
+# Slowdown audit, November 2026
+
+A sweep of `main` (`c02ede3bb`, 13 November 2026) for bugs that freeze the page
+or hold the frame rate down. This page lists what was found, why it happens and
+how to fix it, and names the pull request that carries each fix.
+
+[PERFORMANCE.md](PERFORMANCE.md) is the measured baseline of a healthy build;
+this page is the list of things that fall outside it.
+
+## How it was measured
+
+- Keyless dev server, headless Chrome at 1440 x 900 on Apple silicon (Metal
+  through ANGLE), default globe map.
+- Per scenario: load, a spell at rest, scripted drags and wheel zooms, then a
+  settle. Long tasks were counted with a `PerformanceObserver`, the hot function
+  named from a CPU profile, and `globe.getHeight`, `scene.sampleHeight`,
+  `scene.pick` and friends wrapped to count calls and time.
+- A long task is one main-thread task over 50 ms. With no layers on, the
+  baseline is two at load (longest about 240 ms) and none afterwards, with zero
+  frames rendered while the camera is still.
+- Four profiling sessions ran on one machine at the same time, so absolute
+  timings carry some noise. Ratios against the baseline are what to read.
+
+Nothing here was measured on Google Photorealistic 3D Tiles, on a discrete or
+integrated PC GPU, or with keyed feeds (AISStream, FIRMS, TomTom). Those rows
+are marked "read in code".
+
+## Findings and fixes
+
+Worst first. "Fix" names the pull request that carries the fix.
+
+### Measured
+
+| # | Where | What happens | Fix |
+| ---: | --- | --- | --- |
+| 1 | Bike Share, local data layers, Transit, Rail | 18 s freeze switching Bike Share on over New York (2,519 stations); 2.7 s on Data Centers; multi-second freezes with Transit while zooming | #1688, #1689 |
+| 2 | Shipwrecks | 3 to 5 s freeze when the markers are built over loaded terrain; a share link with the layer on held the loading screen for 22.9 s in one run | #1703 |
+| 3 | Camera | Rendering stops until reload: at a whole-globe view looking straight down, pan a few times, then scroll one notch | #1704 |
+| 4 | Street Traffic | Zooming at 250 m over Manhattan: 25 long tasks in 24 s, longest 746 ms, 6.6 s in total | #1705 |
+| 5 | ALPR cameras | Depth samples inside the overlay paint, repeated on every re-enable: 240 samples and about 1 s across eight toggles. A first-sample freeze of up to 1.9 s was seen in the sweep but not on a second day | #1706 |
+| 6 | Wind | First zoom after load: up to 14 long tasks (longest 414 ms). Re-measured as one-off shader links, not a standing cost: see below | none proposed |
+| 7 | Submarine cables | About 2 s of long tasks on load, heap briefly at 1 GB, about 0.5 s again on each re-enable | #1707 (first step), #1722 (batched primitives) |
+| 8 | Live Flights at whole-globe view with detection labels | Fine on the test machine; at 4x CPU throttle, 8 fps with 143 long tasks in 15 s | #1708 (fleet pass), #1719 (overlay) |
+| 9 | Radio | 7 to 10 long tasks when the directory loads, longest 447 ms | #1709 |
+| 10 | Pointer pick | Not reproduced on a second look: see below | none needed |
+| 11 | CCTV | A still-image camera holds continuous rendering at 60 frames a second although its picture changes about every 10 s | #1710 |
+
+### Read in code
+
+| # | Where | What happens | Fix |
+| ---: | --- | --- | --- |
+| 12 | `/api/firms` | The whole fire set is filtered, serialised and sent uncompressed on every request; 15 to 55 MB estimated in a heavy season | #1711 |
+| 13 | AIS store | Both vessel maps are walked on every position message, and sorted whole once past the cap | #1712 |
+| 14 | Satellite clouds on the 3D Tiles map | A CPU pass over a 4096 x 2048 image, 90 to 180 ms measured on a synthetic canvas, twice in a row | #1713 (canvas path), #1721 (shader) |
+| 15 | NVG and FLIR styles | Over 100 texture reads per pixel, and 60 renders a second while the style is on. Measured afterwards: the reads cost 0.4 ms a frame at 1440 x 900 and 1.2 ms at 2560 x 1440 | #1718 |
+| 16 | Wind server | Full 0.25 degree GRIB fields decoded on the event loop. Measured afterwards: 68 ms blocked per warm refresh, about 250 ms on the first, about 400 ms when all six keys refresh together | #1717 |
+| 17 | Imported geo files and other per-feature ground clamps | Measured afterwards: imported files do not freeze; see below | none needed |
+
+## Cause and fix, per finding
+
+### 1. Terrain height reads
+
+`globe.getHeight` casts a ray against the mesh of the tile under the point, and
+`scene.sampleHeight` renders a depth pass. Transit asked for every trail fix of
+every vehicle several times per poll; Bike Share sampled every station in one
+go. The fix keeps heights per point in one shared reader with a read budget
+(`src/services/terrainSurface.js`) and spreads the depth samples across frames.
+
+### 2. Shipwrecks
+
+`rebuild()` in `src/layers/shipwrecks/index.js` adds 7,958 billboards with
+`heightReference: CLAMP_TO_GROUND`. Each constructor runs
+`Billboard._updateClamping`, which intersects a ray with the terrain mesh,
+synchronously: 4.8 of the 4.9 s. The cost is only paid when terrain is loaded
+at the time the markers are built, so a cold start with the layer in the URL
+sometimes escapes it.
+
+Fix: do not clamp. Wrecks are at sea and the markers already ignore the depth
+test, so place them on the ellipsoid with `HeightReference.NONE`.
+
+### 3. Camera: rendering stops after pan then zoom
+
+Pure Cesium; no application code is involved. Above 1,000 km,
+`ScreenSpaceCameraController` `handleZoom` rotates the camera about the cross
+product of the picked point's normal and the screen centre's normal whenever
+their dot product is between 0 and 1. With the pointer on the exact centre pixel
+the two are the same vector, yet `n · n` often rounds to `0.9999999999999999`,
+so the check passes with a zero axis. `Camera.rotate` then throws
+`DeveloperError: normalized result is not a number` from
+`Scene.initializeFrame`, outside the scene's own try/catch, so
+`scene.renderError` never fires and Cesium stops its render loop for good. The
+pans in the first reproduction only moved the camera to a pose where the
+rounding goes that way.
+
+Fix: make `camera.rotate` ignore a zero or non-finite axis, and recover from any
+stopped loop by restoring the last drawn camera pose and restarting it a bounded
+number of times before showing Cesium's error panel.
+
+### 4. Street Traffic road loading
+
+Three costs in one task after the detail tiles arrive, in
+`src/layers/traffic/source.js`:
+
+- `outsideDetailTiles` clips every coarse z12 road against every loaded detail
+  tile (up to 16) for four regions.
+- The composed set is then handed on with `replace: true`, so
+  `ingestion.js` parses every road again although each tile was parsed as it
+  streamed.
+- `parseRoads` (`model.js`) asks `cachedGroundFloor` once per coordinate, and
+  each call builds several string keys.
+
+Fix: clip once per coarse tile against the detail tiles' union box and slice the
+work across frames (or do it in the vector-tile worker); reuse the roads parsed
+while streaming and parse only the clipped coarse remainder; look the floor up
+once per floor cell within a parse.
+
+This is separate from the one-viewport-tile-at-a-time loading recorded in
+[KNOWN-ISSUES.md](KNOWN-ISSUES.md).
+
+### 5. ALPR cameras
+
+`refineAnchors` and `anchorFor` in `src/layers/alpr/overlay.js` call
+`scene.sampleHeight` from the overlay's paint, four per paint. Each is a depth
+render and a `readPixels`; the first one also compiles the pick shader. Results
+are kept on the entity, so a rebuild or a re-enable samples everything again.
+
+Fix: on a terrain globe read `globe.getHeight` through the shared terrain reader
+instead of sampling; keep the answers in a map keyed by camera that outlives the
+entities; move what sampling remains out of the paint into a queue of one per
+frame.
+
+### 6. Wind
+
+The first reading blamed the layer's 916,476 vertices for slow depth reads
+during zoom. Measuring again did not support that: depth reads cost the same
+with and without Wind, hiding every wind primitive left pick time unchanged, and
+the stalls appear on the first zoom only, as shader programs link (seven
+programs, about 0.5 s, long tasks up to 166 ms). Later zooms had none.
+
+The roughly 30 renders a second are the animation itself. It is already capped
+at 30 Hz and stops when paused, hidden, out of view or under reduced motion;
+lowering the rate would change how the layer looks. `disable()` releases the
+baked geometry on purpose, at the price of one task of about 130 ms on
+re-enable.
+
+A third look, timing every program link on a preview build, settled it: Wind
+owns two shader programs, both link during load in 9 to 18 ms, and none ever
+linked during a zoom. The programs that do link on a zoom are Cesium's own
+globe-surface variants, with Wind on or off, and Cesium links them
+synchronously. No fix is proposed.
+
+### 7. Submarine cables
+
+`src/layers/submarineCables/ingestion.js` loads both collections through
+`Cesium.GeoJsonDataSource` with `clampToGround`. Switching the layer on is one
+task of about 1.4 s, and 1.2 s of it is Cesium reading the terrain height for
+the 1,917 ground-clamped landing markers, all created in the first frame,
+including the ones behind the globe.
+
+Fix, first step: start the landing markers hidden and let the existing horizon
+sweep show the visible ones a few dozen per frame.
+
+Fix, second step: draw from batched primitives in place of the entity data
+sources: one ground polyline primitive for the lines, one billboard collection
+for the landing markers, one point collection and one static polyline primitive
+for the reference stems, all kept while the layer is off. That removes the
+visualiser tasks, the stall on switch-off and on re-enable, and about 900 MB of
+heap. The first switch-on still links shaders and builds the ground-line
+geometry once.
+
+### 8. Live Flights at whole-globe view
+
+Two costs that scale with the 7,500 aircraft of a full poll:
+
+- The fleet pass (`src/layers/flights/rendering.js`, every 80 ms) dead-reckons
+  each visible aircraft and asks `cachedGroundFloor` for it, including aircraft
+  at cruise altitude that can never touch the floor. Each lookup formats numbers
+  to strings several times and rewrites a cache entry: 40% of the busy time.
+- The detection overlay (`src/data/detection.js` `_drawOverlay`) collects every
+  detectable object and strokes every bracket on every frame.
+
+Fix for the first: skip the floor path for a contact clearly above any terrain,
+make the floor-cell key a number, and stop promoting the cache entry on every
+read.
+
+The second turned out smaller than it looked. The bracket strokes are already
+one path per colour, and collecting the objects is about 3 ms of a 56 ms frame
+at 4x throttle once the first fix is in. What costs is the per-object loop in
+`_drawOverlay` (projection, occlusion, alpha, twelve path calls per bracket).
+
+Fix for the second: keep what a sweep built and reuse it on the next paint when
+the view and every contact are exactly what it was built from, compared value
+by value so no layer has to announce a change. Between two fleet passes most
+paints qualify, which cuts frame work by a fifth to a third on a still camera
+at full speed. At 4x throttle a frame outlasts a fleet pass, no paint
+qualifies, and the gain is a few milliseconds; the fleet pass is what
+dominates there.
+
+### 9. Radio
+
+`stationTags()` in `src/layers/radio/categories.js` splits and normalises a
+station's tags on every `hasTag` call, and `buildRadioCategories` runs that for
+every station against every category, twice.
+
+Fix: normalise once per station at ingest and count categories from that.
+
+### 10. Pointer pick
+
+Three sessions saw a single `scene.pick` of 207 to 342 ms during a gesture and
+attributed it to `pickPointer` in `src/voice/pointerPick.js`. Timing every pick
+across pointer rests, drags and wheel bursts did not bear that out: of 78 picks
+with CCTV on over London all but two took 3 to 8 ms, and with five heavy layers
+on, 81 picks took 5 to 30 ms apart from the first (135 ms). `pickPointer` picks
+at most once per pointer rest.
+
+The slow pick is the first pick after a layer appears, once per page load, most
+likely the pick shader compiling, and its stack is CCTV's own hover pick
+(`src/layers/cctv/hover.js`). No fix is proposed: removing the dwell pick would
+only move the one-off cost to the next click.
+
+### 11. CCTV still-image cameras
+
+`src/layers/cctv/projection.js` holds continuous rendering for as long as any
+projected camera is active and redraws the projection every frame.
+
+Fix: hold only for video. For a still image, request one render when a new image
+lands or while a focus fade runs.
+
+### 12. `/api/firms`
+
+`server/providers/firms.js` builds the payload and calls `JSON.stringify` per
+request, with no kept body and no compression, and the client refetches every
+ten minutes.
+
+Fix: keep the response bytes and their gzip and serve those, as the OpenSky
+provider does. The route filters to the trailing 24 hours at serve time, so the
+bytes can only be reused while that filter keeps the same records (about a
+minute at a time): bursts and several clients gain most, and every response is
+about seven times smaller.
+
+### 13. AIS store
+
+`server/providers/vessels/ais-store.js` calls `pruneAisStreamCache()` for every
+message; the static-data map is never pruned.
+
+Fix: prune on a timer, evict in insertion order, and drop a vessel's static
+entry with the vessel.
+
+### 14. Satellite clouds on the 3D Tiles map
+
+`src/layers/weather/infraredImage.js` runs `getImageData`, a per-pixel power
+curve and `putImageData` over the whole canvas.
+
+Fix: a 256-entry lookup table applied in place for the canvas path, which
+halves it; then, for the shell on the 3D Tiles map, the same curve in its
+fragment shader from a strength uniform, so the image is never read back and a
+strength change builds no new image.
+
+### 15. NVG and FLIR styles
+
+`src/styles/surveillance.js` loops 11 x 11 texture reads per pixel and
+`src/styles/thermal.js` totals 106. Timed on the GPU, those loops cost about
+0.4 ms a frame at 1440 x 900 and 1.2 ms at 2560 x 1440 on the test machine:
+real, but small there. The standing cost is that a style with animated grain
+holds continuous rendering at 60 frames a second.
+
+Fix: split the bloom into a row pass and a column pass, which gives the same
+picture from a fifth to a half of the reads, and step the grain's clock at
+15 Hz, asking for one frame per step instead of holding the render loop. The
+grain then changes 15 times a second, which is visible in motion.
+
+Switching on any style stage at all costs more than the bloom did (1.8 ms a
+frame at 1440 x 900 for the one-line noir style). That was not looked into.
+
+### 16. Wind server
+
+`server/providers/wind/decode.js` decodes about a million doubles per message
+synchronously, hourly, for six keys.
+
+Fix: decode on a worker thread, and keep the sampled U and V of the last
+forecast file per model so that choosing an overlay fetches only its field.
+
+### 17. Per-feature ground clamps
+
+`heightReference: CLAMP_TO_GROUND` per feature is also set in
+`installations/rendering.js`, `alpr/presentation.js`, `ghostMaps/index.js`,
+`earthquakes/index.js`, `airQuality/index.js`, `news/index.js` and
+`osmPlaces/index.js`. Their counts were small in these runs (30 to 1,500) and
+none froze.
+
+Imported geo files were the one unbounded case, and measuring them cleared
+them: 20,000 generated points (the import cap) finish in about 1.5 s with a
+longest task near 0.5 s, because `settleGroundPins` in
+`src/data/geoFileImport.js` already places pins with `HeightReference.NONE` and
+resolves heights 256 at a time with a yield between batches. No fix is needed.
+
+## Checked and fine
+
+- Start-up of a production build: viewer ready in about 0.5 s, loading screen
+  gone at 1.6 s, three long tasks totalling about 300 ms.
+- Idle: no frames rendered with no layers on; every render hold was released
+  when its layer went off.
+- Leaks: all 41 layers toggled three times and ten of them eight times; no
+  timer, scene listener, primitive or data source grew per cycle. The heap
+  levelled off near 320 MB over 7.5 minutes of toggling eleven layers.
+- At most one long task after load: Military, Satellites (including 11,903
+  Starlink points), Earthquakes, News, Wildlife, GNSS, Drone Airspace, Military
+  Installations, Fire Perimeters, rain radar, lightning, and satellite clouds on
+  the globe map.
+- Live Flights at city scale with the same 7,500 aircraft loaded: 60 fps, no
+  long tasks.
+
+## Not covered
+
+- Google Photorealistic 3D Tiles: nothing was run there. Depth sampling is the
+  only height source on that map, so findings 4, 5 and 17 may be worse.
+- Live Vessels, FIRMS with data, TomTom traffic, Local ADS-B.
+- Frame rate on anything but one Apple silicon machine.
+- One all-layers run over Manhattan grew the heap from 346 to 638 MB in 60 s;
+  whether that is retained was not established.
+- One Satellites load showed a 5.2 s long task once and never again in four more
+  loads.

docs/PERFORMANCE.md

@@ -6,6 +6,9 @@ specification and should not be used to predict performance on untested systems.
 The original capture artifacts are not included here, so this page records
 results rather than defining a runnable benchmark.
 
+For slowdowns found since, with their causes and fixes, see
+[PERFORMANCE-AUDIT-2026-11.md](PERFORMANCE-AUDIT-2026-11.md).
+
 ## Test context
 
 The baseline was captured on 22 August 2026 with these conditions:

docs/translations/zh-Hant.md

@@ -4,7 +4,7 @@ Catalog: `src/i18n/locales/zh-Hant.json`. General rules: [TRANSLATORS.md](../TRA
 
 `zh-Hant` is the Simplified Chinese catalog (`zh-CN`, credited to abiyeenzo, #914, and EDIT-Aeterna, #979) written in Traditional characters with OpenCC character conversion. Wording and terminology are the same as in `zh-CN`: change a term in [zh-CN.md](./zh-CN.md) and the `zh-CN` catalog first, then convert. Only the script differs.
 
-After `s2tw` conversion, fix the characters OpenCC picks wrongly: 年曆史 → 年歷史, 下遊 → 下游, 並道 → 併道, 發佈 → 發布, 公佈 → 公布, 控制檯 → 控制台, and 周 → 週 where it means a week (1 週/秒).
+After `s2tw` conversion, fix the characters OpenCC picks wrongly: 年曆史 → 年歷史, 下遊 → 下游, 並道 → 併道, 發佈 → 發布, 公佈 → 公布.
 
 ## Style
 

public/models/README.md

@@ -1,33 +1,20 @@
-# Bundled 3D Models
+# Bundled 3D Model Attribution
 
-Every 3D model shipped in this directory is one of the project's procedural
-fleet models under [`fleet/`](fleet/): aircraft, helicopters, ships, buoys,
-transit and rail vehicles, and a few other airborne and ground objects. They
-are generated from code by the project maintainers, contain no third-party
-artwork, and are covered by the repository's [MIT license](../../LICENSE)
-like the source code.
+The model files in this directory are third-party visual assets. They are not
+covered by the repository's MIT source-code license; each remains available
+under the license listed below.
 
-| Path | What it holds |
-|---|---|
-| `fleet/catalog.json` | Every model by name: its file, category, built length, beam and bounds (metres, in Cesium's Z-up model frame: x along the length with the nose toward −X, y across the beam, z up), the clearance below its origin, and its 2D icon class. |
-| `fleet/types.json` | Which model draws which vehicle, and at what length: ICAO aircraft and helicopter type codes, ADS-B emitter categories, AIS ship-type and length ranges, GTFS route types, transit modes and rail categories, with the defaults. |
-| `fleet/aircraft/`, `fleet/helicopters/`, `fleet/ships/`, `fleet/buoys/`, `fleet/transit/`, `fleet/other/` | The models, one GLB each (EXT_meshopt_compression and KHR_mesh_quantization, vertex colours, no textures). |
+| File | Original work and creator | Source | License | Project modifications |
+|---|---|---|---|---|
+| `airplane.glb` | “boeing 747” by [zairiq-123](https://sketchfab.com/zairiq-123) | [Sketchfab model](https://sketchfab.com/3d-models/boeing-747-9b16672038ba48f98e6d80a159044ed9) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Substantially modified and optimized for God's Eye View, including geometry/material simplification and coordinate/orientation preparation. The former 24× runtime calibration is baked into the mesh; location, rotation, and scale are applied; the bounding-box centre is at the origin; and the model uses glTF +Y-up with its nose toward local −X. Geometry remains uncompressed so the aircraft asset does not compete with photogrammetry for Draco worker capacity. |
+| `jet.glb` | “Private Jet” by [Nick the Name](https://sketchfab.com/Nick_The_Name) | [Sketchfab model](https://sketchfab.com/3d-models/private-jet-cbdd1de6ced9461e950eafaa302cc82b) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Repackaged as a glTF binary for the project's military-flight visualization. Imported hierarchy transforms are baked into the meshes; location, rotation, and scale are applied; the bounding-box centre is at the origin; and the model uses meter scale, glTF +Y-up, and the shared nose −X convention. Existing materials are preserved. |
+| `bell206.glb` | “Bell 206 JetRanger” by [terran4627](https://sketchfab.com/terran4627) | [Sketchfab model](https://sketchfab.com/3d-models/bell-206-jetranger-d2f7ba1d671549d4b26aaf834139a1dd) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Optimized for God's Eye View: geometry/material simplification, textures resized to 256 px WebP, orientation/scale vertex-baked to real-world meters (Y-up, nose −X), origin centered. |
+| `c172.glb` | “Cessna 172” by [e737](https://sketchfab.com/e0057537) | [Sketchfab model](https://sketchfab.com/3d-models/cessna-172-64cddaee5aff470682659a8c08525046) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Optimized for God's Eye View: geometry/material simplification, textures resized to 256 px WebP, orientation/scale vertex-baked to real-world meters (Y-up, nose −X), origin centered. |
+| `citation2.glb` | “1990 Cessna Citation, Texture Detailed, Exterior” by [BlenderCommunityHead](https://sketchfab.com/aboodgoudagad) | [Sketchfab model](https://sketchfab.com/3d-models/1990-cessna-citation-texture-detailed-exterior-a78839624fe64900a8352cb23462350a) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Optimized for God's Eye View: geometry/material simplification, textures resized to 256 px WebP, orientation/scale vertex-baked to real-world meters (Y-up, nose −X), origin centered. |
+| `mq9.glb` | “MQ-9” by [IProZenoN](https://sketchfab.com/IProZenoN) | [Sketchfab model](https://sketchfab.com/3d-models/mq-9-fabe963feb354c5584b51f9c470c3f7e) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Optimized for God's Eye View: geometry/material simplification, textures resized to 256 px WebP, orientation/scale vertex-baked to real-world meters (Y-up, nose −X), origin centered. |
+| `b789.glb` | “Boeing 787-9” by [Nobilis 2](https://sketchfab.com/nobilishornet2) | [Sketchfab model](https://sketchfab.com/3d-models/boeing-787-9-b6711e2e698e4e469675c1154a50b7a3) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Optimized for God's Eye View: geometry/material simplification, textures resized to 256 px WebP, orientation/scale vertex-baked to real-world meters (Y-up, nose −X), origin centered. |
+| `atr72.glb` | “ATR 72 - 600” by [Oyan3D](https://sketchfab.com/oyan3D) | [Sketchfab model](https://sketchfab.com/3d-models/atr-72-600-1e1a7186f7444d288675262fcee44744) | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) | Optimized for God's Eye View: geometry/material simplification, textures removed with dominant material colors baked into PBR factors (flat “abstracted” style), orientation/scale vertex-baked to real-world meters (Y-up, nose −X), origin centered. |
 
-How the layers use them:
-
-- Each layer draws a vehicle with the model its type selects, scaled to the
-  type's published length (a vessel to its reported AIS length and beam), so
-  every model is drawn life size. Ground objects, balloons and obstacles in
-  the `other` category are drawn at the size they are built.
-- An aircraft model's origin is its axis; it stands that model's scaled
-  clearance above the ground. A ship's origin is its waterline, and a transit
-  or rail vehicle's the road or rail top (a monorail or gondola is raised by
-  its `transit.elevation` entry, for a source that names those modes).
-- Transit and rail models carry a `_TINT` vertex attribute that marks where
-  the route or category colour is painted.
-- The whole set loads once, up front, when a layer that draws it starts; no
-  model is fetched while a session runs.
-
-`src/data/fleetModels.js` reads the set and chooses each vehicle's model;
-`src/data/fleetModels.test.mjs` checks that every model `types.json` names is
-in the catalog and on disk, and that no GLB exceeds 3 MB.
+CC BY 4.0 permits sharing and adaptation, including commercial use, provided
+appropriate credit is retained, the license is linked, and modifications are
+identified. These credits do not imply endorsement by the original creators.