The data was literally duplicated, it was primarily textures and models. The reasoning was to reduce loading times for players with harddrives because the seek head wouldn't have to move as much to find the data it was looking for, but because the textures and models are streamed in at the same time as the level was being generated (a slow process regardless of the storage medium) the duplicate data wasn't actually helping that much.
Unsure if you couldn't deduplicate it lol, maybe Steam would complain because your files are altered?
I've been thinking about doing this exact thing too but prices for the BC250 have gone up like 50% in a few months. A lot of ebay listings are now people selling build kits with 3D printed cases and dubious power supplies bundled in for an even more inflated price.
Still significantly cheaper than a Steam Machine but if you're thinking about pulling the trigger on a BC250 you should do it relatively quick before Youtubers start making videos about the idea (again) and everything gets more expensive (again) lmao.