parsePlanVideosJson rebuilt each video from a fixed field list that had no
playbackRate, so chunk workers indexed extracted frames at rate 1 while the
planner mixed audio at the authored rate. A data-playback-rate="1.5" clip
showed its picture at 1x in planV2/renderChunkV2/assembleV2 output, while the
single-process render was correct.
Carry playbackRate through the parser, accepting a finite number or a rate
lane, and pin it with a JSON round-trip test that checks the chunk-side frame
lookup for 1.5x, 0.5x and a rate lane.
* fix(gcp-cloud-run): run the Cloud Run image on Bun 1.4.2 so remote downloads cannot hang
* fix(engine): make the download deadline abort the body pipeline
Residual of #3340: runtime-assigned src is skipped by the static parse, so
the browser snapshot was still keying clips by author id. Colliding scenes
collapsed onto one window.
Parent src-only scans skipped multi-format <video>/<audio> markup, so those
elements were never extracted, downloaded, or mixed and rendered blank/silent.
Lint now accepts a child <source src> as a resolvable media src.
compileTimingAttrs/injectDurations used parseFloat, so data-start="intro"
wrote a NaN data-end and extract preferred that over duration; parseNumeric
now skips the id-ref (parseVideoElements already resolves it).
collectRenderMedia's resolveHostWindow likewise read host data-start with
parseFloat, so chained sub-composition slots (data-start="hook") stacked at
0-2s and every scene after the first rendered black. It now resolves host
starts through the shared resolveReferencedStart, matching the media parsers.
Fixes#3361.
Seek the GSAP timeline once per timestep and sample every auto-start
video, instead of re-walking per media element. Keeps probe cost
proportional to duration, not video count.
Browser probe end reflects full source duration, not the scene slot.
Never extend a compile-time end — only fill missing or shrink — so long
recordings sliced across short scenes don't inflate extraction windows
and time out black on Cloud Run (ARC-13403).