Regression Test Plan

Manual checklist for validating MTL Explorer end-to-end. Most checks use the app UI. Fixture setup, service-state changes, and API evidence need access to the disposable test environment.

For each item, perform the action and confirm the expected result. Note any deviation, including blank screens, error toasts, slow responses, or visually broken layouts.

Fresh-server runs do not use a personal Google account. They check Google Health disabled, unconfigured, empty, and demo states only. Connected Health checks are a separate local pass with the owner's existing account; see Google Health scope. Do not copy account credentials, tokens, local health data, or private tracks to a test server or report assets.

Coverage Accounting

Coverage IDs such as TRD_01 are stable report references. Once added, an ID must never be renamed, renumbered, reused, or changed to mean a different check. Add new checks with the next unused number in the closest matching ID prefix. If a check becomes obsolete, keep the ID and mark it NOT APPLICABLE or explain the replacement. Packet results and full-regression reports must reference coverage IDs, not section numbers alone.

• ACC_01 Treat every checklist bullet as a required coverage item unless it is explicitly not applicable to the run.
• ACC_02 Do not mark a section PASS when any bullet inside it was skipped, only spot-checked, or verified indirectly. Mark the section PARTIAL, NOT COVERED, NOT APPLICABLE, BLOCKED, or FAIL as appropriate.
• ACC_03 In full-regression reports, include enough coverage detail to show which bullets were actually exercised and which were not. Broad section summaries are acceptable only when all child bullets have supporting evidence.
• ACC_04 Capture compact screenshots for representative working user-facing functions as well as failures, so reports provide a visual overview and not only defect evidence.
• ACC_05 If time, tooling, viewport, data, permissions, or environment constraints prevent a check, record that constraint explicitly instead of silently collapsing the check into a parent row.

Required Data-Change Regression

Run this in every full regression and every release-candidate data-change pass.

Public Test Data

• DAT_01 Use at least five public internet GPX files with real track sequences (trk / trkseg / trkpt). Waypoint-only files (wpt without trkpt) are not valid positive import evidence.

• DAT_02 Prefer GPX files with timestamped trackpoints so duration, speed, moving time, and period statistics can be verified.

• DAT_03 Record for every source file: source URL, source page/license note, destination filename, SHA-256, byte size, trkpt count, timestamp count, imported track id(s), and imported track name(s).

• DAT_04 Suggested verified GPX source: https://github.com/gps-touring/sample-gpx with raw files such as:

  • https://raw.githubusercontent.com/gps-touring/sample-gpx/master/BrittanyJura/JuraRoute72011.gpx
  • https://raw.githubusercontent.com/gps-touring/sample-gpx/master/BrittanyJura/MoselradwegAusWiki.gpx
  • https://raw.githubusercontent.com/gps-touring/sample-gpx/master/BrittanyJura/Vitry-le-Francois_Langres.gpx
  • https://raw.githubusercontent.com/gps-touring/sample-gpx/master/BrittanyJura/VoieVerteHauteVosges.gpx
  • https://raw.githubusercontent.com/gps-touring/sample-gpx/master/RoscoffCoastal/Lannion_Plestin_parcours24.4RE.gpx

• DAT_05 Use at least one public FIT activity file with GPS positions. Suggested verified FIT source: Garmin's official FIT SDK examples, e.g. https://raw.githubusercontent.com/garmin/fit-javascript-sdk/main/test/data/Activity.fit.

• DAT_06 Do not count non-GPS FIT files or waypoint-only GPX files as positive evidence. They are useful negative tests only: MTL Explorer should fail, ignore, or mark them clearly without adding map tracks, corrupting stats, or blanking the UI.

• DAT_07 For measure / comparison / virtual-race checks, include at least one repeatable two-point segment with two or more tracks crossing the same two zones. Use public tracks or fully synthetic anonymized GPX data; do not use private local GPX files as fixtures or report evidence.

• DAT_08 Use the fully synthetic, non-private media set generated by the packaged /app/demo/generate_regression_photos.py helper. It uses the same shared renderer as the demo photos and ffmpeg for short videos. The helper produces four GPS JPEGs, two camera-time-only JPEGs, one H.264/AAC MP4 with embedded GPS and capture time, one H.264/AAC MOV with capture time but no GPS, and a matching six-point synthetic GPX activity. Record the generated manifest, including checksums and ffprobe data. Copy the GPX into data/gpx/<run-id>/ and all eight media files into data/media/<run-id>/; keep the manifest and deletion backups outside the watched media tree.

• DAT_09 Prepare fully synthetic GPS cases for timing and motion: repeated visits to two ordered zones, a recorded pause below one minute, a longer observed stop, a source recording break, and missing/repeated/reversed timestamps. Record expected intervals and geometry independently of displayed results. Keep these fixtures outside the five-GPX count assertions and retain them through segment/replay checks. Do not derive them from local tracks.

Five-GPX Import, Index, Map, And Stats Flow

• IMP_01 Capture baseline map count, track-browser count, statistics totals, data-freshness token, and GPS indexer status.
• IMP_02 Import the five GPX files through the Settings → Import & sync upload UI or by placing them in the documented watched import folder.
• IMP_03 Wait for indexing to finish. If live file watching does not react, trigger Rescan GPS from Settings → Maintenance and record that it was needed.
• IMP_04 Confirm upload/index status: all five source files reach completed state, no unexpected GPS index failures appear, data freshness changes, and background jobs (Duplicate Finder, Exploration Score) settle.
• IMP_05 Reload from the freshness banner or helper reload action → the map, track browser, filters, and statistics all show the new data.
• IMP_06 Verify each imported file by name: it appears in track browser search, on the map, in statistics summaries, and in at least one filter result.
• IMP_07 On the map, zoom to the imported tracks, click each track, verify selection/detail opening, point popups, visible line geometry, and no stale or duplicated lines.
• IMP_08 In statistics, verify count increased by five unless a source file legitimately split into multiple tracks; if it split, record the source-to-track mapping and expected count.
• IMP_09 Verify totals changed in the correct direction: total distance, duration, ascent/descent, activity breakdown, period charts, rankings, heatmap density, and track-browser summary row.

Delete-Two-Track Flow

• DEL_01 Delete two of the imported source files from the watched import/upload folder. If the test environment only exposes browser upload and not the watched folder, mark deletion sync BLOCKED and run this flow in the install/full-regression environment.
• DEL_02 Wait for automatic delete processing or trigger Rescan GPS.
• DEL_03 Verify the two deleted tracks disappear from the map, track browser, filter results, selection lists, heatmap, related-track lists, and statistics totals.
• DEL_04 Verify the remaining imported tracks still display and open correctly.
• DEL_05 Deleted-track API probes or stale deleted-track URLs are not pass/fail criteria for this deletion flow. The frontend regression requirement is that deleted tracks no longer appear in user-visible map, browser, filter, heatmap, related-track, detail, or statistics surfaces.

FIT Conversion Flow

• FIT_01 Import the FIT activity file with GPS positions.
• FIT_02 Verify it is accepted by upload/import, indexed successfully, displayed on the map, searchable in the browser, and included in statistics.
• FIT_03 Open the FIT-backed track details → overview, graphs, quality, events, related tracks, mini-map, and point popups render as they do for GPX-backed tracks.
• FIT_04 Download original source file → the downloaded file remains FIT and matches the uploaded checksum.
• FIT_05 Download as GPX → a valid GPX file downloads and contains real trkpt trackpoints, not only waypoints.
• FIT_06 If GPSBabel or FIT conversion is unavailable, the UI shows a clear conversion/indexing error and the failure is recorded as blocking for FIT support.

Other Supported Track Formats

• FMT_01 The server accepts .gpx, .fit, .tcx, .kml, .kmz, .igc, .nmea, .geojson, and .gdb. For a full release regression, test at least one GPS-bearing sample for each available format, or mark that format NOT COVERED with the reason.
• FMT_02 For each non-GPX format tested, verify upload acceptance, GPSBabel conversion, map display, details/charts, statistics inclusion, Download original source file, and Download as GPX.

Media Fixture Setup Flow

• MED_06 Capture the media count, MEDIA indexer status, and media freshness baseline. Import the DAT_08 matching GPX first and wait for GPS indexing. Place all six generated JPEGs and both generated videos in the watched media folder, open Settings → Maintenance, trigger Rescan Media, and wait for MEDIA indexing and correlation to settle. Accept the freshness reload and verify all eight files are completed without failures. Preserve the generated manifest and exact source files outside the watched tree for later deletion evidence and recovery.

1. Sign-in And First Load

• SGN_01 Open the app while signed out → you are redirected to the login screen.
• SGN_02 Sign in with valid credentials → you reach the map.
• SGN_03 Sign in with wrong credentials → a clear error appears and you stay on login.
• SGN_04 If demo mode is active, the login screen shows the demo credentials banner.
• SGN_05 Sign out → you return to login; signing in again works.
• SGN_06 The splash screen (logo, background, message) displays during startup and disappears once the map and tracks are loaded.
• SGN_07 If startup fails (e.g. server down), a retry is offered instead of a frozen splash.
• SGN_08 "MTL Explorer" branding appears in About / public-facing copy.
• SGN_09 Browser back/forward navigation between views works without errors.
• SGN_10 On desktop and phone, in light and dark themes, Sign In retains its translucent background, blur, white label, and visible keyboard focus. Check hover where supported, original field height and blur, and demo banner readability; successful and failed sign-in retain the same treatment on return.

2. Map And Tracks

• MAP_01 Base map and overlays load on first open.

• MAP_02 All your tracks appear on the map; the total/visible count is correct.

• MAP_03 Newly imported tracks from the required data-change flow appear without a full browser restart after accepting the freshness/reload prompt.

• MAP_04 Deleted tracks from the required data-change flow disappear from all map sources, selection lists, and popups.

• MAP_05 Zoom in on a track → detail/precision improves (no duplicate or broken lines).

• MAP_06 Fast pan/zoom doesn't leave stale lines, missing tiles, or runaway loading spinners.

• MAP_07 Direction arrows appear on tracks at high zoom when Track Points & Direction is enabled. Use a track whose current high-zoom rendered shape has visible in-viewport point vertices. A sparse two-point or straight synthetic track whose endpoints are outside the viewport is not valid evidence: the line may cross the screen while no arrow marker is expected at the visible midpoint.

• MAP_08 Click a single track → it highlights and details open.

• MAP_09 Click an area where several tracks overlap → a selection list appears; picking one opens its details. Categories whose lookups complete with no results have no chip or result group. Pending and failed/cancelled categories stay visible. If a selected pending category completes empty, All becomes active and shows the remaining results or “Nothing found here”. GPS details open expanded without widening the standard desktop panel. They use a separate foreground sheet; mobile details fill the screen to the top. The dispatcher keeps a constant height while details open, resize and close. Back, Close and browser Back reveal its unchanged size and category/page.

• MAP_10 Deselect / close the selection → the map returns to its normal state.

• MAP_11 Clicking a rendered track-point marker shows a popup with the expected metrics (time, speed, elevation, etc.). Click an actual direction-arrow/point marker, not only the connecting track line.

• MAP_12 Swiss Mobility routes popup (where applicable) shows nearby official routes and closes cleanly.

• MAP_13 Intentional remote raster mode: with mtl.map-server.tile-mode=remote, /api/map/config exposes remoteRasterStyles for light, light-topo, and dark, and does not expose legacy remoteTileUrl. Select OSM Light, OSM Topo Light, and OSM Dark; each loads map tiles from its configured provider URL, shows the matching attribution, keeps the map interactive, and makes no /api/map-proxy tile requests.

• MAP_14 Remote raster fallback from local vector mode: when local vector PMTiles are not ready or fail at runtime, the map does not blank or freeze. It switches to the configured remote raster style for the selected map theme, uses the configured attribution, and continues to support pan, zoom, track display, and track selection. If the environment cannot safely simulate local tile unavailability, mark this row BLOCKED with the missing control.

• MAP_15 Manual remote raster source override: in a local-vector deployment, open Map → Map style, switch Map Source from Auto to Remote, and verify the base map reloads from the configured remote raster provider without /api/map-proxy tile requests. OSM raster themes remain selectable, Swiss vector themes are not offered in Remote mode, the setting persists after reload, and Reset restores Auto.

• MAP_16 Open the Map overview, then Map style, terrain, personal data, and route-overlay screens. Each Back action restores the overview; summaries reflect the changes. Repeat on a phone and after reload; reset restores defaults without changing the UI theme.

3. Track Details

• TRD_01 Open at least one GPX-backed track and one FIT-backed track from user-facing navigation (map, browser, or stats) and record the track ids/source filenames.

• TRD_02 Opening a track loads its overview, charts, related-tracks list, event list, mini-map, and quality info.

• TRD_03 Switch between Overview, Graphs, Quality, Related, and Events; tabs do not refetch in a loop, lose state, or show blank panels.

• TRD_04 Elevation, speed, distance, and gain charts render with readable values.

• TRD_05 Graph controls work: time/distance x-axis toggle, range band toggle, point-count slider, and graph-height slider update charts without layout breakage.

• TRD_06 Hovering a chart highlights the matching point on the mini-map and hovering the mini-map highlights the chart. No stale cursors remain after leaving either surface.

• TRD_07 The track shape preview (small thumbnail) is visible in browser, filters, stats, related tracks, and selection lists.

• TRD_08 Download original source file (GPX/FIT/etc.) → file downloads and matches the uploaded one.

• TRD_09 Download as GPX → a valid GPX file downloads even if the source was FIT or another format.

• TRD_10 Change activity type (e.g. hiking → cycling) → saves successfully; energy/calorie values update automatically.

• TRD_11 Energy "what-if" recalculation (custom rider weight, etc.) → updates the displayed values without permanently saving.

• TRD_12 Exclude from statistics toggle → the track stops counting in stats overview; re-including it brings it back.

• TRD_13 Related tracks show duplicates and previous/next tracks; clicking one navigates to it.

• TRD_14 Events tab shows detected stops / GPS gaps where present; selecting an event highlights the matching mini-map position and deselects cleanly.

• TRD_15 Open Track Details from Statistics → Tracks and from Filter → Review tracks. Closing details returns to the same originating sheet and list state. Repeat with browser Back/Forward and on narrow mobile; a direct track-detail link closes to the map. The referring panel stays open at a constant height underneath GPS details. Check the latest-activity shortcut and Related track links too; returning must reveal the existing panel without a closing or reopening animation.

• TRD_16 Open Quality → GPS review on the synthetic continuity cases. Original and processed paths, source quality, processing decisions, and unsupported intervals remain distinguishable. Selecting a review item highlights its local geometry without connecting disconnected recording spans.

• TRD_17 On the synthetic motion cases, compare Overview and Events: moving + stopped + unknown equals elapsed duration for a complete timeline. Short observed pauses count toward stopped time but do not create break events below one minute. Longer breaks show approximate boundaries; missing or invalid time remains unavailable/unknown, not a zero-duration stop. See motion rules.

• TRD_18 Local connected Health pass only: compare whole-track and moving heart-rate summaries with their server values. Whole-track heart rate includes recorded stops; moving heart rate uses moving intervals only. Time/distance graphs share the full track extent and preserve partial coverage. A track without samples shows unavailable values and a useful sync-status link. On fresh servers mark NOT APPLICABLE: personal Health account required.

• TRD_19 Open a track without Health samples on the fresh server. Core metrics, graphs, mini-map, Quality, and Events remain usable; missing heart rate does not cause repeated requests, blank core charts, or fabricated zero values.

4. Filters

• FLT_01 Open the filter panel → previously saved filter is still active and shown as a chip.

• FLT_02 Browse the filter catalog; search and grouping work.

• FLT_03 Choose a filter in Filter view, confirm Use this view, and open Criteria (required missing criteria open automatically). Valid parameter edits auto-apply immediately; clearing or removing parameters resets their effect; active chips, visible count, map, legend, and stats reflect the current state without stale pending UI.

• FLT_04 Date, text, and geo parameters all save and re-apply correctly after reload.

• FLT_05 Geo drawing: draw a circle, rectangle, and polygon; undo, cancel, finish, and clear all work; saved shapes reappear next time.

• FLT_06 Applied filter updates: visible track count, map colors, legend, and stats — without a full page reload.

• FLT_07 Legend reflects the active filter (categories or gradient); collapsing/hiding groups updates the map immediately.

• FLT_08 Clearing the filter restores all tracks.

• FLT_09 Choose a grouped year filter, select two result categories, and verify the map, visible count, track browser, heatmap, Statistics Overview, Trends, and Stats Tracks all use the same selected tracks.

• FLT_10 Repeat result-category selection with main activity groups and exact activity types such as Walking and Hiking. Labels and counts remain correct.

• FLT_11 Change date, text, or geo parameters while an exact result-category selection is active. The selection remains active; newly discovered categories remain unchecked.

• FLT_12 Select none and verify the empty result is intentional and stable. Select every available category and verify it returns to All categories when there are no unavailable saved categories.

• FLT_13 Change parameters so a selected category has no matches. It remains visible as unavailable and can be removed.

• FLT_14 Switch to a different filter and verify the result-category selection clears. Select the current filter again and verify its selection remains.

• FLT_15 Reload with an exact result-category selection. It is restored before the first resolution and all filter-aware views still match.

• FLT_16 Hide a group with the map legend visibility control. Map visibility changes, but statistics do not. Changing the global Result categories selection resets temporary legend hiding during map reload.

• FLT_17 NOT APPLICABLE in the current UI: automatic first-visit filter guidance was removed. Keep this ID for historical reports; the current overview and on-demand guidance are covered by FLT_18 and FLT_22.

• FLT_18 Reopen Filter as a returning user on desktop and narrow mobile. Guidance stays inside the result summary, remains compact without overflow, and More info opens the full explanation without the Important badge. Back returns to Filter; Close closes the sheet.

• FLT_19 Pause and resume from the Filtering on / Filtering off switch in the result summary. The result status, map, statistics, persisted state, and switch position remain synchronized on desktop and mobile; no header switch or duplicate action appears.

• FLT_20 Open Review tracks from the shared footer beside Show on map. The review uses the same track-browser view as Statistics → Tracks, including search, summary, responsive table/cards, sorting, pagination, map selection, and track details. Recheck Statistics → Tracks after the shared-component change.

• FLT_21 Filter and Review tracks use the standard Filter sheet dimensions and detents. Verify the nested review does not jump to an oversized desktop sheet and remains usable at both mobile detents.

• FLT_22 Open Filter as a new and returning user. The overview provides Filter view, Criteria, Included categories, Map colors, Review tracks, and Show on map. Open each available inner screen and return with Back. Closing or Show on map keeps the filter active; reopening from the map result count works.

• FLT_23 Stage a different Filter view. Compatible criteria are retained; criteria that would be cleared are listed before confirmation. A custom category order shows a notice that it will reset to the new view's default. Cancel preserves the old view and order. Use this view applies the choice, clears old category selection and custom order, and opens missing required criteria without sending an invalid query.

• FLT_24 In Criteria, exercise Last 30 days, This year, Last year, and All time. Both date bounds update together. A reversed range shows a field error, disables Done, and retains the last successful map result; correcting it recovers. While paused, valid edits retain their draft and Apply and enable restores the intended result.

• FLT_25 Search Included categories, stage a subset, and verify the displayed draft count before Apply while the map remains unchanged. Cancel discards edits; Apply updates map, review, Statistics, heatmap, and Segments consistently. Empty search and zero selected categories provide recovery actions.

• FLT_26 Reset a non-default view with criteria, categories, and colors, then Undo while offered. The previous configuration and result return. Map color changes and Show all for hidden map groups leave the statistical track set unchanged. SQL inspection remains a secondary action.

• FLT_27 On a narrow phone, complete and cancel circle, rectangle, and polygon drawing from Criteria. Return to Criteria with the correct retained/updated values. Back, Done, Review tracks, and Show on map remain reachable without horizontal scrolling.

• FLT_28 In a view with a preferred palette, apply No coloring and reload. The choice survives without changing dates, categories, or matches. Repeat with filtering paused, then resume and switch between category and gradient views. Reset uses the view defaults; Undo restores No coloring. Repeat on desktop and phone.

5. Track Browser And Statistics

• TBS_01 Track browser lists all (or filtered) tracks with name, date, distance, duration, activity, etc.

• TBS_02 Search matches names, descriptions, dates, distances, durations, activity, and file paths.

• TBS_03 Sort by each column works; summary row reflects what is currently visible.

• TBS_04 Quick-view/preset buttons switch the browser subset correctly and preserve usable sorting/search behavior.

• TBS_05 Clicking a row opens the track's details.

• TBS_06 Statistics overview shows total distance, time, elevation, activity breakdown, rankings, milestones, and period charts.

• TBS_07 Stats are correct for: empty dataset, a single track, and many tracks.

• TBS_08 Stats update after the required five-GPX import and again after deleting two imported tracks; no stale deleted-track totals remain.

• TBS_09 Time-period charts (daily/weekly/monthly) render and switch correctly.

• TBS_10 Clicking a stats entry navigates / filters / highlights as expected.

• TBS_11 Highlight drilldowns open the expected track list, open a selected track, and expose excluded-highlight counts where applicable.

• TBS_12 Statistics use the same resolved track set as the active filter: apply a geo-drawn filter (circle, rectangle, or polygon) that changes the visible map count, then verify Stats Overview totals, Trends track totals, and the Stats Tracks tab all match the filtered map count. Repeat after a reload to confirm the fallback ID-resolution path still matches.

• TBS_13 When Statistics Overview shows Showing X of Y tracks, activate the summary with pointer and keyboard. It opens Filter directly on desktop and mobile.

• TBS_14 Open Statistics → Trends → Charts. Media is the final chart; All media is selected by default in the control at the top. Both choices expose visible, keyboard-accessible help. The choice is shared with Map settings → Your data and Overview media counts.

• TBS_15 Change grouping and sub-unit. Activity and media periods remain aligned, with zero values where needed. Apply date, category, geographic and custom filters: media follows the earliest valid start and latest valid end of the actual matching tracks, including gaps and both endpoints. Older, newer and undated media never appear. Empty/undated track selections show no media. All media includes unlinked items; Track related additionally requires a selected link to a matching track. Map markers refresh with both filter and scope changes, and stale requests cannot restore old markers.

• TBS_16 Select a non-empty photo or video stack. The period mosaic uses the same track-defined range and opens with the correct total, All/Photos/Videos filters, stable newest-first cards, 60-item Load more pages, recoverable loading/error states, and shared viewer navigation. In Track related scope, Open activity returns to that activity's Media tab; the action is absent in All media scope. There is no undated drill-down.

• TBS_17 Open Statistics and each Overview, Trends, and Tracks entry. Back returns to the hub and restores focus; track search, sorting, and Trends controls survive the round trip. The latest-activity preview respects the current track set, opens details, and handles an empty set.

• TBS_18 Change Trends chart height and order, reload, and restore defaults. The saved layout returns, height stays usable within the viewport, and the media chart remains included. Repeat on a narrow phone in light and dark themes.

6. Planner

• PLN_01 Open Planner → Start a new plan (or Continue your current plan when a draft exists); pick a routing profile (e.g. hiking or touring bike).

• PLN_02 Click on the map to add waypoints → a route is computed and drawn.

• PLN_03 Insert a waypoint on an existing leg by dragging the route.

• PLN_04 Move and delete waypoints; clear, undo, and redo all work.

• PLN_05 Live stats bar (distance, ascent, time) updates as you edit.

• PLN_06 Elevation profile renders and hovering it highlights the matching map point.

• PLN_07 Save plan, list saved plans, load a saved plan, delete a plan.

• PLN_08 Download plan as GPX → file is valid and matches the planned route.

• PLN_09 If the routing engine (BRouter) is missing data for an area, the UI shows a clear "segment downloading / unavailable" state instead of an unhandled error.

• PLN_10 Existing planned routes still display even when the planner has trouble fetching new data.

• PLN_11 Touch dragging on mobile works for placing and moving waypoints.

• PLN_12 From the Planner hub, open the contextual guide, return, draw a draft, and return to the hub again. The draft survives; map clicks edit only in the drawing screen. Saved plans and drawing remain reachable on desktop and mobile.

• PLN_13 Select numbered waypoints and exercise tap-to-move, cancel move, insertion, removal, reverse, and trim before/after. Undo/redo restores geometry, stats, and routing profile. Drag release must not add an extra point; point actions remain reachable on touch screens.

• PLN_14 Search and open a saved plan. Its exact saved geometry, profile, notes, and stats load and the map frames it above the panel. With an unsaved draft, cancel replacement and confirm the draft survives; then explicitly replace it. Loading failures offer retry without silently clearing the draft.

• PLN_15 Save after an edit while routing is pending. Save waits for current geometry and prevents duplicate submission. A failed save keeps name/notes for retry. Editing a saved plan offers Save copy and leaves the original intact; success offers a working GPX download. A backend 200 alone is not evidence of browser file delivery.

• PLN_16 With unsaved changes, attempt to leave/reload the page and cancel the browser prompt. The draft remains. Trigger a slow route request, then change profile, clear, undo, or load a saved plan; a late reply must not overwrite the newer state.

• PLN_17 On desktop and phone, use More map, Show tools, slow and fast panel drags, and grabber taps. Totals, profile, and Undo stay visible without scrolling; the profile stays readable and its map cursor aligned until the totals-only size. Downward flicks never close the drawing panel, the chosen size returns on the next visit, and route updates keep the previous totals until the new result arrives.

7. Measuring And Comparison Tools

• MCT_01 Open Segments, place start and end zones, and choose Analyze → crossing attempts appear with speed/time/distance.

• MCT_02 Clicking a result opens that track's details / segment view.

• MCT_03 Stop the measure tool → all temporary markers and listeners are cleaned up.

• MCT_04 Segment comparison: pick several tracks → comparison chart + map align them correctly even with missing data.

• MCT_05 Sub-track / segment extraction (between two points on a track) returns the expected slice.

• MCT_06 Segment geometry sanity: after selecting a measured segment, the comparison map line stays within the selected tracks' real local bounds. There must be no straight global line, jump to [0,0], or off-continent segment such as a line toward South Africa.

• MCT_07 Place multiple ordered zones, adjust the radius, undo a zone, and Analyze. Zone counts and the results describe the same filtered tracks. Empty results explain how to adjust the selection; editing zones invalidates the old analysis.

• MCT_08 Using repeated-visit fixtures, select individual attempts and a reference, then switch among result table, charts, and Replay. Labels, colors, reference, and exact zone/visit identities stay consistent; Select all is explicit. Consolidated summaries retain separate valid visit counts for each metric.

• MCT_09 Compare exact attempt boundaries, distance, elapsed time, energy, ascent/descent, and power with server results for the synthetic fixtures. Changing chart density changes presentation only. Fractional boundary work must not include whole-track energy outside the segment. See segment rules.

• MCT_10 Analyze missing/repeated/reversed-time and recording-gap cases. Untimed attempts remain inspectable where geometry permits; unavailable timing is labelled rather than ranked as zero. Core charts preserve gaps and supported partial coverage without inventing a bridge.

• MCT_11 On a fixture with a recorded stop, compare finish faster/slower labels and Time difference at equal traveled distance. The latter stays on a distance axis, includes stopped time, ends at shared coverage, and is not described as spatial lead/lag.

• MCT_12 Inspect aligned graphs by hover, click-to-pin, keyboard, and touch. The shared value panel and map position agree; selecting an attempt from the whole value card updates highlighting. The panel sits beside charts when space permits and below them on a phone.

• MCT_13 Local connected Health pass only: verify segment HR, coverage, cardiac cost, and unavailable reasons against server values for matching local samples. Ordinary HR includes recorded stops; cardiac cost uses covered moving intervals including coasting. Consolidated HR uses covered-time weighting. On fresh servers mark NOT APPLICABLE: personal Health account required.

• MCT_14 On a fresh server without Health data, selected core results, charts, and replay still load. Missing Health results remain unavailable. Retry a failed core attempt without losing other selected results; rapidly changing selection must not restore stale responses.

• MCT_15 Reimport or remove a disposable synthetic source after discovery, or let its attempt reference expire. Loading the old attempt offers Refresh results. Refresh rediscovers current data instead of combining old geometry with new metrics. Preserve the shared fixture set needed by later checks.

8. Animation And Virtual Race

• AVR_01 Open Animate on desktop and mobile: existing tracks remain visible until preview or playback starts; Play collapses to the map-first transport bar; pause, resume, stop, expand, reset, date range, and speed controls work; estimated duration updates.

• AVR_02 Virtual race: multiple racers move together; ranking and racer cards update in real time.

• AVR_03 Stopping or finishing animation/race leaves map gestures and tools usable (no stuck state).

• AVR_04 Virtual race GPS geometry regression: start a race from a measured segment with multiple racers and verify each racer marker and trail stays on the actual segment. The map must not zoom to world-scale bounds, draw a long straight line away from the route, or show a racer/trail near [0,0] or South Africa.

• AVR_05 Start Replay from selected segment attempts. Play/pause, speed, reset, and elapsed-time scrubbing update markers, trails, and rankings consistently. Rankings retain activity/date labels; switching back to charts preserves selection and reference.

• AVR_06 Replay a synthetic attempt with a recording gap and one without usable timing. Unsupported portions must not become invented movement; timing-unavailable attempts are not presented as valid timed racers. Stop and return to the map with normal gestures restored.

9. Media (Photos and Videos)

• MED_01 With no saved map preference, the media layer is enabled and photo pins appear by default. Disable it and reload to verify it remains disabled; enable it and reload to verify it remains enabled.
• MED_02 Pan/zoom → media is loaded for the current viewport (not the whole world at once).
• MED_03 Click a pin → photo preview opens; next/previous navigation works.
• MED_04 HEIC photos display correctly (converted server-side).
• MED_05 A missing/broken photo shows a recoverable error, not a blank sheet.

Activity Photo Correlation Flow

Use a timestamped public or fully synthetic activity and synthetic photos. The fixtures must cover embedded EXIF GPS, camera time without GPS, a saved camera correction, and an overlapping activity window.

• MED_13 Open Track Details → Media. The baseline request returns persisted timeline rows in capture-time order. Reload the page and restart the server; the same rows, positions, origins, and order remain.
• MED_14 Verify an embedded-GPS photo is labeled Photo GPS. Read-only database inspection shows its original EXIF point unchanged and the resolved origin EXIF_EMBEDDED.
• MED_15 Verify a camera-time-only photo is labeled Estimated and its marker lies on the activity route. Database inspection shows a persisted TRACK_INTERPOLATED correlation and a separate resolved projection; it must not create original EXIF coordinates.
• MED_16 Enter a camera offset and choose Preview. The UI marks the result as unsaved, changes only camera-clock photos, and leaves EXIF GPS time unchanged. Reset returns to the persisted baseline without a database change.
• MED_17 Save the previewed correction. Reload and restart to verify the correction remains. Use Clear clock correction and verify the original capture metadata is unchanged and the baseline correlation is restored.
• MED_18 Set a manual location and optional note. The item and marker are labeled Set by you. Read-only inspection confirms the manual, EXIF, and track-derived positions remain separate and resolution precedence is USER_ASSIGNED > EXIF_EMBEDDED > TRACK_INTERPOLATED. Clear the assignment and verify it falls back to the preserved EXIF or track position.
• MED_19 With two eligible activities covering one photo time, verify the UI shows Ambiguous, stores all eligible correlations, and selects the same deterministic winner after recalculation and restart.
• MED_20 On a disposable installation, add, replace, and delete relevant activity data. Verify only photos in the old or new activity time window are queued, persisted correlations are recalculated, and resolved map positions update without a full media-library scan.
• MED_21 With at least 100,000 synthetic media rows and 300 activity rows, record EXPLAIN ANALYZE and HTTP timings. Map bounds must use the spatially indexed resolved-position projection; a Track Details timeline must use the indexed selected-correlation lookup. No request may interpolate across the full media library at runtime.
• MED_22 Assign a manual location, then preview a non-zero camera offset. The card, viewer, and mini-map keep the Set by you position and clear action while route and capture-time preview data updates underneath it.
• MED_23 Exercise a correction whose saved offset has non-zero seconds, such as 3603. The compact label is valid (+1h, not +1.h). If position provenance is temporarily unavailable, the UI says Position unknown and never claims Photo GPS.
• MED_24 After setting or clearing a manual location, request the same map bounds again without waiting. The marker updates immediately and the response uses Cache-Control: no-store.
• MED_25 In a disposable PostGIS test, delete a correlated activity through the production deletion service. Its photos are requeued, the queue row survives deletion, and recalculation falls back to another activity, EXIF, or no resolved position.
• MED_26 Seed one deterministically failing correlation item followed by a healthy item. The failing item is deferred with attempt/error metadata and the healthy item completes; it must not retry the blocked batch forever.
• MED_27 On the main map, click a cluster containing multiple synthetic photos. A chooser distinguishes This cluster, Current map view, and Photos along a GPS track when a GPS activity passes the clicked location. Open the only activity directly. With several activities, verify the standard map track chooser is reused and the selected row opens its Media tab. Verify each destination and visible scope label. The cluster viewer opens at most 200 members without changing map zoom or camera; previous/next lazily loads the adjacent bounded page. After manually zooming until markers separate, click a single marker and verify the same chooser distinguishes This photo from Current map view. Repeat the complete flow on desktop and at 390 × 760.
• MED_28 Open the matching synthetic activity's Media tab. Timeline rows and mini-map markers are server-paged in capture-time order: 100 items by default, no more than 200 when requested, and only the current page is rendered. Exercise first, middle, and last pages on the MED_21 100,000-row dataset and confirm page navigation remains responsive without an unbounded response or DOM.
• MED_29 In the activity mini-map and viewer location map, Photo GPS, Estimated, Set by you, and unknown positions use the same circular camera symbol; only color and accessible text communicate provenance. No origin uses a different geometric marker shape. Clearing a manual position restores the previous color and label.
• MED_30 On desktop, open a six-photo activity viewer and test the large side previous/next controls, keyboard arrows, swipe, direct Nearby selection, Nearby collapse/expand, and image zoom/pan/reset. Navigation controls must not zoom the image. Details can hide and return; Taken includes local date and time plus the GPS-time or camera-clock source. The location map stays in the lower-right details area, fits the full selected activity plus photo automatically, exposes only standard zoom in/out controls, and Open on main map closes the viewer at the photo location.
• MED_31 At 390 × 760 and 375 × 667, the viewer opens photo-first with Details closed. Details opens as a native vertically scrollable dark sheet and reaches metadata, location map, main-map action, and download. Nearby collapses to one compact dark row retaining filename and count; the phone label shortens to Nearby. Expanded and collapsed states reach the bottom edge with no white safe-area or sheet-padding strip.
• MED_32 Compare the separate viewer window controls. Panel maximize fills the MTL Explorer viewport; the adjacent fullscreen action enters browser fullscreen for the entire viewer and clears panel maximize. Details and Nearby retain their current states and can each be changed in either mode. Previous/next and zoom still work. The same fullscreen action or Escape returns to the normal still-open viewer; restoring panel size does not change Details or Nearby.
• MED_33 Give the viewer a 100,000-ID selection centered away from either end. It renders only a bounded 200-thumbnail window, keeps the current item/count correct, supports direct nearby selection, and shifts the window in both directions while the DOM count stays bounded. Repeat with a paged map cluster and activity: reaching either strip or previous/next boundary loads the adjacent 200-photo page. Record DOM counts, requests and interaction timings; do not treat a screenshot alone as performance evidence.
• MED_34 With the application in both light and dark themes, open map, activity and statistics photo viewers with no stored viewer preference. Each defaults to dark. Use the header shortcut to select light and verify the header, photo stage, navigation, Details, metadata dividers, mini-map area, Nearby dock, thumbnails, controls, focus states and mobile details sheet all adapt with readable contrast. Reload and open each viewer entry point; light persists only for photo viewers while the application theme remains unchanged. Switch back to dark and repeat maximize, browser fullscreen, Details, Nearby, zoom and previous/next interactions.
• MED_35 Open an activity Media tab on desktop and at 390 × 760. The timeline, position origins, paging and selection are visible immediately; Camera clock, Adjust locations, and per-photo edit actions are not. Open the single Photo tools disclosure, verify the camera preview is usable, then enable Adjust locations and verify edit actions appear. Closing Photo tools also closes an open editor and leaves the timeline uncluttered. Repeat first/next/last paging on the 100,000-photo fixture and verify no page renders more than the selected limit.

Video Playback and Location Flow

Use both generated DAT_08 videos. The MP4 is the mandatory native-playback case. The MOV verifies a second container and a camera-time-only location path.

• MED_36 Verify both generated videos finish MEDIA indexing and appear on the map, in the matching activity timeline, in media navigation, and in the photo/video statistics counts. Every thumbnail surface returns a JPEG poster with a visible video/play indicator; no surface shows a broken image.
• MED_37 Open mtl-regression-video-gps.mp4 in the viewer. Native playback is required: metadata loads, duration is positive, dimensions are non-zero, play advances current time, and pause, seek, resume, keyboard/swipe navigation, poster, Details, Nearby, and original download work. Verify the full response is 200, a byte-range response is 206 with correct Content-Range and length, an unsatisfiable range is 416, and the downloaded original matches the manifest checksum.
• MED_38 Open mtl-regression-video-estimated.mov and record canPlayType('video/quicktime; codecs="avc1.42E01E, mp4a.40.2"'). When the original plays, require the same play, pause, seek, and resume checks as MED_37, even if that support is broader than expected. When actual decoding reports an unsupported source or decode error, require a clear explanation, an Auto/480p/720p/1080p quality selector, and a user-started compatible stream. Preparing, progress, slow-conversion guidance, cancellation, failure, and retry states remain usable while the poster and checksum-matching original download stay available. The compatible HLS stream must start, play through its full duration, and retain pause/resume behavior. A blank viewer or broken poster fails either path. Verify 200, 206, and 416 serving behavior for the original MOV as well.
• MED_39 Verify video location provenance end to end. The MP4 retains its embedded coordinates and resolves as EXIF_EMBEDDED even when the matching activity could supply an interpolated point. The MOV has no original coordinates, resolves onto the matching GPX route as TRACK_INTERPOLATED, and is labeled Estimated. Set a manual MOV location and note, verify USER_ASSIGNED wins without overwriting the track-derived point, then clear it and verify the persisted track-interpolated location returns. Compare UI values with the manifest and read-only database rows.
• MED_40 Reload the app and restart the server. Reopen both videos from the main map and activity timeline. Poster, navigation, native or compatible playback, locations, origins, range requests, and downloads remain stable, with no failed MEDIA rows, blank viewer, or uncaught console error.
• MED_41 Generate a short disposable archival-codec/PCM MOV with ffmpeg in the test run's temporary fixture directory; do not commit the video. Follow video-transcoding/README.md if the test browser plays the archival codec natively. Exercise the full compatible flow at 480p and 720p. Reopen during conversion and after a page reload: the same media revision and quality reuses the active or retained session instead of starting a second ffmpeg process, and switching quality starts only the newly selected bounded profile.
• MED_42 With short test-only limits, verify one active conversion, one retained completed result, global and per-session byte caps, maximum runtime, disconnect grace, completed expiry, oldest-completed eviction, and startup orphan cleanup. Active output is never evicted. Crossing a hard limit stops ffmpeg, removes the session directory, and gives the viewer a recoverable message. Record the temporary directory size and process/session counts before, during, and after cleanup.

Delete-Two-Photo Flow

MTL Explorer currently has no media-delete control. Remove only the disposable source fixtures from the watched folder; perform the rescan, status, freshness, and map verification through the GUI.

• MED_07 Before deletion, record the filenames, media IDs, checksums, coordinates, visible pins, MEDIA indexer status, and freshness token for mtl-regression-delete-a.jpg and mtl-regression-delete-b.jpg. Capture a compact map screenshot showing the disposable photo pins.

• MED_08 Remove those two source files from the watched media folder. Only manipulate the disposable DAT_08 fixtures. Keep exact backups outside the watched tree until the flow is complete.

• MED_09 In Admin, trigger Rescan Media and wait for MEDIA indexing to settle. The removed count increases by two, no new failure appears, and a media freshness change is available. Reload through the freshness action.

• MED_10 Return to the recorded viewport. The two deleted photo pins are absent, cannot be opened through map navigation, and the two remaining synthetic photos still display and navigate correctly.

• MED_11 Pan away and back, zoom across cluster levels, hard-reload the app, and revisit the same viewport immediately. The deleted pins do not reappear, the remaining pins are not duplicated, and the bounds response is not cached.

• MED_12 On the disposable installation, use read-only database inspection to verify that each deleted source retains an indexed_file row with status REMOVED, has a DELETE snapshot in media_file_audit, and has no active media_file row. Record only synthetic fixture data in the evidence.

• MED_43 Open a public or synthetic HEIC/HEIF image and a resized JPEG preview, including a format requiring fallback where available. Verify decodable output, orientation, dimensions, viewer recovery, and original download. Capture libvips/ImageMagick fallback logs or response headers when exercised; use the media fixture and verification guide.

• MED_44 Request several uncached image previews while the media process limit is occupied. Busy responses remain recoverable; retry eventually displays the image without an endless broken-preview state. Use disposable media and record the controlled contention setup.

10. Heatmap And Overlays

• HMO_01 Toggle the heatmap → it draws over the map without hiding the tracks; respects opacity.
• HMO_02 Toggle each map overlay (Swiss / OSM / etc.) independently; opacity sliders work; ordering above/below tracks stays correct.
• HMO_03 After changing filters, the heatmap updates accordingly.

11. GPS Location

• GPS_01 Browser geolocation requires a secure origin. In quick-install runs opened from a remote plain-HTTP host, mark live GPS permission/marker checks NOT APPLICABLE - expected browser limitation; test those rows on localhost or HTTPS.
• GPS_02 Enable GPS → permission prompt; on accept, the locate marker appears.
• GPS_03 "Follow me" mode keeps the map centered until you pan away (drifted state).
• GPS_04 Permission denied / disabled state shows a clear message.
• GPS_05 Disabling GPS removes the marker and stops updates.

12. Location Search

• SRC_01 Open the search → type a place name → results appear.
• SRC_02 Pick a result → map flies to it and a marker is placed.
• SRC_03 Clear search / pick a different tool → marker is removed cleanly.
• SRC_04 Empty / no-result queries show a clear message.

13. Globe Mode

• GLB_01 Zoom out far enough → globe view engages automatically.
• GLB_02 Zoom in → map returns to flat view.
• GLB_03 Manual disable of globe is respected (does not auto-re-enable until you re-enable it).
• GLB_04 Zoom limits don't trap the map at edges.

14. Settings And Operations

• ADM_01 Open Settings; the overview and grouped section navigation are reachable and usable on desktop and mobile.

• ADM_02 Track file upload: drag or pick GPX/FIT/etc.; upload availability, accepted formats, progress, success, unsupported-format errors, and empty-file errors are clear.

• ADM_03 Indexer status: shows GPS and media pending/running/completed/failed/removed state; refresh updates over time.

• ADM_04 Manual rescan: Rescan GPS and Rescan Media show queued/already-running/not-ready states without breaking map interaction.

• ADM_05 Background jobs: Duplicate Finder and Exploration Score progress is visible and settles after imports.

• ADM_06 Operational tasks: vector map tiles, location search, and routing segment status show ready/downloading/unavailable/disabled states with useful detail.

• ADM_07 Data freshness: shows last-update timestamp and offers reload.

• ADM_08 Server log: log lines load and refresh.

• ADM_09 About and credits: the public About page shows the expected map, library, and data sources before and after login; Back returns to the previous Settings section or the login/map fallback.

• ADM_10 Garmin export tools (if present): status of installed exporters; install/update actions report success or error.

• ADM_11 Closing/reopening the dialog doesn't lose state mid-action.

• ADM_12 Direct Settings section URLs, browser Back/Forward, mobile Back, and closing the sheet keep the section and map routes synchronized.

• ADM_13 Settings opens with Preferences in the first group and compact expandable system status. Status links open the correct detail section. Desktop navigation remains beside content; phone Back stays outside scrolling details. Section headings receive keyboard focus.

• ADM_14 Import & sync shows compact Garmin and Google Health entries. Google Health Manage/setup and Processing links keep their own return path in Settings. The server-log error state offers retry; long log lines and service URLs do not force page overflow.

• ADM_15 In a disposable browser context, compare credentials-only sign-out with full local-data cleanup from Session. Confirmation explains the scope and Cancel preserves the session. Confirmed cleanup removes the selected local state without deleting server tracks or saved plans. Never use this check to delete the owner's Health mirror.

15. Data Updates And Sync

• SYN_01 After server-side data changes (new import, re-index, or restart), a data-freshness banner appears.

• SYN_02 Reloading from the banner refreshes cached tracks and stats.

• SYN_03 The required five-GPX import and delete-two-track flow passes: indexer state, freshness banner, map, browser, stats, filters, heatmap, and details all reflect the new source-of-truth files.

• SYN_04 FIT conversion import changes freshness and cache state the same way a native GPX import does.

• SYN_05 Map data warnings start compact. Expanding shows details and actions. Minimize collapses the details while keeping the warning chip visible; reopening retains Reload map and Data status or Processing details. Verify desktop and phone, including a status update while minimized.

• SYN_06 Logging out and back in does not re-trigger an automatic data refresh repeatedly.

• SYN_07 Indexer-running state surfaces as a badge but doesn't block map interaction.

• SYN_08 On the disposable server, stop the app, add/replace/delete synthetic GPS and media source files, then restart. Startup catch-up updates source states and dependent results without a manual rescan. Accept reload and verify map, browser, media, and statistics. Record each change and restore the baseline; retain fixtures needed by later checks.

• SYN_09 Open Settings Processing while indexers and other status consumers are active. Shared status polling remains bounded rather than multiplying per panel; closing consumers stops unnecessary polling. A failed status section exposes its error while available sections and map interaction continue.

16. Appearance (Theme And Map Style)

• APP_01 Switch between light and dark mode → the whole UI re-themes immediately (text, panels, dialogs, sheets, dropdowns, tooltips, charts).
• APP_02 No text is unreadable (white-on-white or black-on-black) in either theme.
• APP_03 Charts re-color on theme switch without needing a reload.
• APP_04 Selected theme persists across reload and login.
• APP_05 Hard refresh in dark mode does not flash the light theme first.
• APP_06 Map theme is independent: each of the available map styles (light, dark, grayscale, light-topo, topo-contrast, swisstopo, swisstopo-color) can be selected with either UI theme.
• APP_07 Selected map style persists across reload.
• APP_08 Layer opacity sliders, basemap dimming, and reset-to-defaults all behave and persist.
• APP_09 Open the User Guide and a topic link in each app theme. Navigation, search, empty results, tables, and code blocks follow the saved preference on desktop and phone. Reload retains it; an open guide updates when the app theme changes in another tab. Printing uses the light palette. From the navigation, About, Filter, and Planner, browse several guide pages, then use Back to app. The guide tab closes and the original app keeps its map position, filters, open tool, and unsaved planner draft without reloading. A guide opened directly still offers a working app link.

17. Locale, Units, And Formatting

• LOC_01 Numbers, distances, durations, and dates render in the expected locale format.
• LOC_02 Changing locale (if available) updates formatting across the app without reload artifacts.
• LOC_03 Locale persists across reload.
• LOC_04 Boundary values (zero, very large, negative gain, null elevation) render sensibly, not as "NaN" or blank.
• LOC_05 Metric / imperial measurement flow: in Settings → Preferences → Region & units, select Metric and record representative values from Statistics, Track Browser, Track Details and graphs, a map point popup, filters with numeric distance/elevation parameters, measuring or segment comparison, Planner, and replay/animation. Switch to Imperial (US) without changing the underlying track, route, or segment. Distance, elevation, speed, vertical rate, and weight values must convert consistently (km/m/km/h/m/h/kg → mi/ft/mph/ft/h/lb), while duration, dates, watts, and watt-hours remain semantically unchanged. No checked view may mix metric and imperial units. Reload the app and confirm the imperial preference and converted values persist, then use Use default or select Metric and confirm metric values return after another reload. Capture compact paired evidence from Preferences, Statistics, and at least one detailed metric surface.

18. Responsive / Mobile / Touch

• MOB_01 Test at a narrow mobile width and with touch input enabled.

• MOB_02 Bottom sheets and the navigation sheet drag, snap, and close correctly.

• MOB_03 Tables, charts, and map controls stay usable; no text overflows.

• MOB_04 Planner waypoints can be tapped, dragged, and inserted with touch.

• MOB_05 Map gestures (pinch, double-tap, drag) work after using each tool.

• MOB_06 NOT APPLICABLE in the current UI: the Filters/Settings tab switch was replaced by focused screens. Keep this ID for historical reports; use FLT_22-FLT_27 for the replacement flow.

• MOB_07 The phone toolbar exposes GPS, Map, Filter, Planner, and More. More opens Statistics, animation, Segments, Health, and Settings. Desktop exposes its direct tools with Settings separated at the bottom. Switching tools and returning closes obsolete sheets and retains the correct route.

• MOB_08 Open short and long Settings, Filter, Statistics, and Planner sheets at narrow phone, tablet, short desktop, and wide desktop sizes. Small height growth may remove unnecessary scrolling with one opening animation; long content remains scrollable. Manual sizing takes priority until reopening, and focus/touch targets remain usable.

19. Offline And Network Issues

• NET_01 Installed PWA / installed web-app mode only: after installing MTL Explorer in the browser and loading once online, reload while offline → cached tracks and tiles still display. A normal browser tab is not expected to pass this offline reload check; mark that row NOT APPLICABLE or NOT COVERED unless the app is installed as a web app.
• NET_02 A flaky connection shows recoverable error states, not a blank screen.
• NET_03 A 401 response on a protected route redirects to login. A 403 response reports the failed request while preserving the valid session; normal responses recover without signing in again.
• NET_04 Service worker update: a "new version available" prompt appears after an update; accepting it reloads cleanly.
• NET_05 Cached basemap with no network in an ordinary browser tab: sign in online with the automatic/local vector source and wait for the low-zoom map archive and tracks to be saved. Block all backend and external network requests, then pan/zoom within the saved archive's zoom range into an area not previously viewed; open Statistics and return to Map without reloading the document. Cached geography and tracks still render, with a clear cached/offline state and usable map controls. Unavailable fonts, sprites, overlays, or terrain must not leave the basemap blank or startup waiting indefinitely. Restore connectivity: online maps recover without clearing storage or signing in again, and the selected filter and map-source preference remain intact. Capture cache preparation, blocked requests, and before/offline/reconnected screenshots on desktop and a narrow viewport. Full offline document reload remains NET_01.

20. Error Recovery

• ERR_01 Trigger or simulate: failed track load, failed map config, failed media, failed planner route, expired session. Each shows an actionable message (retry, re-login, dismiss) rather than freezing or going blank.
• ERR_02 Rapid switching between tools does not leave the previous tool's markers, listeners, or cursors behind.

21. UX Performance

• UXP_01 Start from a signed-in, warmed desktop session with the required imported data present and background jobs settled. Capture browser main-thread responsiveness and every completed first-party application API fetch/XHR while running three consecutive journeys through map pan/zoom, Filter apply/reset, Statistics Overview and Tracks, track search and details tabs, Planner open/close, and return to the map. Each input must show visible feedback within 200 ms, no measured main-thread stall may exceed 500 ms, every first-party application API response must finish within 2 seconds with its expected status and UI result, and no request may remain unexpectedly pending. Exclude static assets, map tiles/providers, uploads, downloads, and intentionally cancelled requests from the API budget. After the third journey, map pan/zoom and the main navigation must still respond. Save a timing table with action feedback, maximum stall, API route/status/duration, pending or failed requests, and any console error; a screenshot alone is not enough evidence. Any exceeded budget, unexpected response/result, error, or loss of interaction fails this coverage item.

22. Google Health

Scope and prerequisites

A fresh-server regression can pass without connected Health coverage. State that scope in the report; it does not certify synchronization or real Health records. On local runs, record missing data or a required account action as BLOCKED with the prerequisite. Do not change or seed the owner's mirror to manufacture boundary cases. Existing synthetic unit/integration tests can support those cases, but do not count as live account regression evidence. Capture redacted UI/status evidence only; no tokens, account identifiers, private record payloads, routes, or health values in committed assets.

Fresh-server checks

• GHL_01 With MTL_GOOGLE_HEALTH_ENABLED=false, MTL Explorer starts and the normal sign-in, map, filter, browser, statistics, and Settings smoke flow works. Google Health shows a disabled state; connection/sync actions fail without a 5xx response, and no Google jobs, remote requests, route files, or mirror rows are created. Local-mirror cleanup is a separate guarded operation, not a sync action.
• GHL_02 With Google Health enabled but no client credentials, the same smoke flow works. The UI shows setup is required, connection and sync remain unavailable, and no remote sync starts. Restore the original disposable app configuration afterward.
• GHL_03 Local account lifecycle check only, with separate explicit authorization: connect/reauthorize, inspect partial consent, sync, browse Health and exercise scope, disconnect, and delete the local mirror as applicable to the authorized action. Progress separates the current step from cumulative records and older-history coverage. Disconnect retains the mirror; authorized deletion requires its confirmation and must preserve ordinary imports and saved plans. On fresh servers mark NOT APPLICABLE; routine local passes exclude lifecycle mutations. Record the exact authorized scope and any unexecuted parts rather than claiming the full lifecycle passed.
• GHL_04 On an unconfigured fresh server, open Settings → Import & sync → Google Health → Start setup. The assistant explains required account/project setup, read-only access, Desktop OAuth configuration, and manual return-address paste. Copyable environment/Compose examples use placeholders. Back and Close work; stop before account authorization.
• GHL_05 With demo mode enabled and synthetic health disabled, Health shows its non-interactive preview and server Health data access is blocked. With mtl.demo-health.enabled=true, Overview, Data & Records, and matching track heart rate show labelled synthetic data through the ordinary APIs; generation progress is visible in Settings → Processing. In both cases account connection and Google synchronization remain blocked, and no Google remote requests run. Restore the original app configuration afterward.
• GHL_06 Without a local mirror, open Health Overview and Data & Records. Overview keeps empty chart periods and setup links usable without requesting status or catalogue counts. Data & Records shows disabled/unconfigured/empty states and offers retry for failed local status requests. No endless spinner or effect on ordinary map/statistics use.

Local connected pass

• GHL_07 With the existing connected account, open Health → Overview and Data & Records, then Back. Navigation and focus return to the hub. Manage and sync notices open the appropriate Settings section and preserve a useful return path on desktop and phone.
• GHL_08 Use date presets, including Today, then custom ranges. Presets apply immediately; custom Cancel preserves the previous range and Apply uses the selected inclusive calendar dates. Test desktop and single-calendar phone layouts; invalid/over-limit ranges explain the correction.
• Opening Overview requests charts only, including after closing and reopening it. It does not request or poll full Google Health status, type summaries, sources, or raw records. Expanding Latest values requests summaries and then its measurements; closing during loading prevents follow-up reads. Reopening a successfully loaded section reuses its values until a shared sync update or screen closure. Sync updates leave collapsed latest values unloaded.
• GHL_09 Switch All, Activity, Heart & vitals, Sleep, and Body with available local records. Auto/Day/Week/Month/Year grouping follows calendar boundaries, Monday-start weeks, and the displayed time zone. Intraday uses one local day and the same grouped API; select a day, move to adjacent days, and open a day from a daily measurement chart with See day in detail. Use the 6 h/1 h/15 min/1 min presets and the earlier/later arrows, set the exact range 12:00–12:15 and Apply, and drag or pinch on a graph; every graph shows the same window without another request, and expanding and returning keeps it. Select a point and verify the marked time on every graph, each graph's nearest reading, presets centered on that time, exact timestamps, averages, observed ranges, line breaks only at long gaps, and the empty-window note. Full day restores the whole day. Daily-only values stay unavailable at Intraday. Returning to calendar grouping restores the previous range. Latest values remain independent of the Trends range. A period with no records is empty, not a fabricated zero measurement.
• GHL_10 Switch summary/daily variation and observed-range choices where available; reload and verify each chart's preference. Hover, pin, keyboard-select, and touch-select a period. Coverage, dated extrema, clipped periods, and in-progress labels agree with the returned server data. Expanded charts preserve grouping and selection; Back/Escape restores focus and scroll.
• GHL_11 Where existing local data supplies overnight sleep, naps, or daylight-saving boundaries, confirm each whole session belongs to its start date in the viewing time zone. Daily stacks and longer-period daily averages use the server's session duration. Record unavailable cases explicitly; do not insert synthetic health records into this account.
• GHL_12 In Data & Records, apply type, representation, source, and date filters, load another page, open Details, reset, and retry a failed request if one occurs. Rows match the selected filters without duplicates. Date-only records retain their calendar date. Test responsive cards/table and unknown/null values when present in local data.
• GHL_13 Observe an ordinary background sync. Current pass/window progress stays distinct from cumulative mirror counts; the verified snapshot age remains visible during work and counts refresh after the cycle stops. Pending history and retry ranges link to Processing. Do not force a reconnect, full-history sync, failure, or cancellation on the owner's account merely to satisfy this row.
• GHL_14 When an existing exercise has an imported route, open the associated track and exercise-only scope. The matching route is shown, ordinary statistics stay consistent, and normal repeated background sync does not duplicate it. Record missing eligible routes as a local data prerequisite.
• GHL_15 Switch metric/US customary units and light/dark themes across Health, track HR, and segment HR/cardiac cost. Distances, weight, and speed convert where applicable; bpm, watts, and beats/kJ retain their meaning. Coverage/unavailable labels and charts stay readable at phone width. Restore the owner's original preferences.

Synthetic synchronization acceptance

Use an isolated database and synthetic provider responses for these checks. Do not reload, reset, or mutate the owner's connected mirror.

• Fast work runs first, full work advances across bounded cycles and restart, and busy ticks create no replay. Show incomplete fast coverage under overload.
• Named corrections and timestamp moves retain one current identity; differing nameless observations survive. Invalid/conflicting pages cannot replace records.
• Complete children sweep only their own bounds/stamps, including NULL stamps; empty days finalize, revoked owners fail, and rollback leaves no final mark.
• Empty batches write daily audits, stop at gaps/caps, and shrink when data appears. A 1970 floor with parked gaps remains responsive through cached status.
• Range/all reload supersedes matching finals without pre-deletion. Unrequested parked portions remain parked. Retry releases failures; reset clears every new state.
• Exercise removal retains files, owners, and associations. Reappearance and track replacement reuse ownership. Adding location permission exports stored routes.
• Persist Retry-After cooldown; count all HTTP attempts, probes, retries, and decoded response bytes. Record real-account envelope, density, delay, and cost measurements separately before making production performance claims.

Suggested Regression Passes