I have a theory that I call "just one small fix". I think FOSS media players can handle more video files (not just formats) because most of the times the errors that makes them unplayable are just small mistakes in the way things are encoded.
Does a certain camera produces only even timestamps? Let's check for that fix it, it is "just one small fix: one if and a division". Is that MPEG encoded with swapped croma? Let's detect it and fix it, it is "just ome small fix: one if and a swap".
Over the time FOSS projects accumulate all these fixes and one day you realize that they can read practically everything as if a single person sat down and wrote that super-robust code in one go. Proprietary software cannot benefit from this kind of drive-by contributions.