# Scenarios page video export and matching simulation
Date: 2026-10-14
Scope: research_domains/lrigit/scenarios.html, the existing SVG flywheel. This is the page explicitly selected by the user.

## Product behavior
- A seventh selectable matching scenario now explains atomic requirements, resource recall, primary evidence, capability gaps, candidate/work-package review, V1/V2 comparison and cross-challenge reuse.
- It is visibly fictional simulation, with candidate A/B and explicit unverified gaps. It does not claim live matching, successful cooperation, mail sending or TableGit writes.
- Original six scenes and the default insight scene remain.
- Scenario selectors now fit two desktop rows / three mobile rows. Lower content and video aspect ratio shift together.
- The export section is on scenarios.html itself: current step (from its beginning) or current scenario's full 12-step cycle.
- Export takes an immutable snapshot of scenario, phase, language, layout, selected report version and speed. Later page changes cannot change an in-flight export.
- The existing speed button cycles 1x, 1.5x, 2x, 4x, 0.25x, 0.5x.
- A complete cycle has source duration 42.7 seconds. Output duration is source duration / selected speed.
- Output: local WebM/VP8, 24fps, same display aspect ratio; progress, cancellation and cleanup. No external upload.
- Frame materialization reuses flywheel-view's scene renderer and shared visualState calculations used by the live player.
- Existing simple horizontal SMIL particles are sampled to static coordinates. Unrecognized animation fails explicitly rather than silently disappearing.
- The video encoder is shared with the earlier independent TableGit scene export; the TableGit page is not the target of this change.

## Actual verification
- Local Node syntax checks: PASS.
- Pure renderer test: PASS for 7 scenarios x 12 phases x 2 layouts x 2 locales = 336 deterministic frames, plus prompt interpolation, ranges, matching artifact label, even output dimensions and preservation of the default scene.
- VP8/WebM constant-frame container fixtures decoded by ffprobe:
  1x: 42.700000 sec / 1025 frames.
  1.5x: 28.466666 sec / 684 frames.
  2x: 21.350000 sec / 513 frames.
  These validate media container/timing including the >30-second cluster boundary; they are not browser SVG captures.
- Shared Shell validation was PASS before the seventh-scene addition; markup continues to use the same shell/style contract.
- Configured formal topic-workflow gate was PASS before the seventh-scene addition and is rerun by formal publication.
- Full repository validation: FAIL at the unchanged meaningful_missing=34 style-coverage baseline.
- Current browser playback/download and SVG-to-WebCodecs integration: NOT_RUN. Previously observed local Chromium absence / cloud navigation block remain; no real-browser acceptance is claimed.
- Existing validate-flywheel.mjs now expects seven scenes; local pure export/render checks are distinct from its repository integration reads.
- Layout source/code checks do not certify live font metrics or viewport interaction.

## Follow-up
Browser acceptance must download current-step and full matching-scenario videos at slow/normal/fast rates, verify raster frame content and duration, cancel a long export, check desktop/mobile font bounds, and switch scene/locale during an in-flight immutable export.
Do not equate container fixtures with browser export acceptance or interpret fictional matching examples as authoritative matching records.