In our documentation, we use "streaming" to refer to the UI of giving the user random access into a media file. It's true that supporting that properly does mean that there is an API that allows random access into the byte stream. I emphasize the distinction because our PDF viewer (the one that's embedded in WA) also uses random access into the data stream, but we don't refer to it as "streaming a PDF".
These API's aren't documented, but range requests are part of the generic HTTP spec. So it's likely to be more stable than our other undocumented API's.