I think the pronunciation, specifically the blending of the end of “upside” and beginning of “down”, turns it into one of those compound words that your brain interprets as an independent word, rather than a combination of its composite parts.
Unused to wonder if the radio announcers that are always reciting the station call letters found that the letters stopped sounding like individual sounds, and the whole recitation became a sort of “word” for them. Like “You’re listening to 102.9FM WBLM!” Did it stop being “double-you bee ell emm,” and turn into more of a mashup of “dubbleyabeeyelmm”?
True, the difference is pretty subtle, especially to a listener, but I wonder strange things sometimes…