The proprietary program has this blog post: https://www.cleverfiles.com/help/advanced-camera-recovery-in...

I think the key point is that cameras don't write the video files in one long contiguous block on disk. They internally split it up and write in an interleaved fashion. It even mentions low-level tricks like manipulating the FAT table so the moov atom which is written last appears at the beginning of the file.

I've found when you have a file with stops and starts, it's because the extraction process is not familiar with how the data is laid down on the storage media. So it sees 'I have a file'...and if it's better it sees 'the name of the file is here' and then 'it's this big' and then 'here's the linked list of clusters for that file'....or it starts at the first cluster and gets as far as it can before it runs off the tracks.