When the parent talks about jumping to random byte offsets, they mean you don't have the first part of the file at all. You just have an arbitrary 512-MB chunk out of the middle.

But they were claiming that a single monolithic file would break too, which is not the case. The browser does a range request to get the first part, then a range request to get the part you're playing, and it works.

Right. Like tuning into a digital-cable signal "in the middle"†. You just get bytes of the stream starting from an arbitrary offset, without having seen/processed anything before that (and without even being able to request anything before that), and you need to resynchronize from what you've got.

† I mean, a digital-cable video stream is always "in the middle" unless you're just starting a VOD stream, but still.