Our seeks use nsBuiltinDecoderReader::DecodeToTarget() to decode up to the seek target after the main seek returns. This pumps the decode until the decode produces a VideoData and/or SoundData which contains the seek target. But for audio we'll begin playback at the start of the SoundData, not at the actual seek target. There could be an unknown amount of sound data before the seek target, reducing the accuracy of our seek.
Created attachment 544668 [details] [diff] [review] Patch v1: Prune unnecessary leading audio samples