My (hugely overthought, but I’ve typed it now) theory is that it’s mimetic, and that it depends on how we learn to read & write – and that it’s specifically because we don’t in this country tend to hammer home spelling and grammar as a series of rigid rules, beyond the basics.
Most people will concentrate enough on writing at school to pick up those basic rules, which is why most people know to start sentences with a capital letter, or (taking it up a level) that ‘i’ generally comes before ‘e’.
But beyond that we don’t really learn ‘rules’ as such – we just kind of pick things up by osmosis, and (in part because English spelling is irregular) learning specific spellings or specific distinctions becomes a matter of trial-and-error. But the amount of trial-and-error you get depends on how much reading & writing you do, and that actually varies widely.
That’s where the loose/lose problem comes in. If your education has included a certain amount of reading & writing, that trial-and-error will at some point have encountered the distinction between ’loose’ and ’lose’, and you’ll never make that mistake again.
But evidently huge numbers of people don’t quite get enough to encounter that distinction. Instead, what happens is that their writing is a mix of correct application of the actual rules, supplemented by guesswork or what-looks-right for things where they don’t know the actual rules.
That’s generally enough for daily life – if you’re essentially copying and adapting writing that’s already out there, you’ll get most things right even if you’ve never encountered them as rules.
However, and particularly now that social media is a major source of reading, it also means that the pool of what-looks-right includes a lot of other people’s guesswork. Hence there’s scope for errors like loose/lose – if they’re both common, and you’ve never learnt the difference, and there’s no obvious reason to prefer one over the other, then it’s basically 50/50 which one you pick.
And what’s more, now you’ve added one more example to the pool of what-looks-right. Soon enough we’ve reached a position where, absent a mass learning event, there’s no way to get rid of it – eventually the mistake will be so routine that ‘loose’ will just be accepted as an alternative spelling of ‘lose’, and you’ll feel slightly daft for feeling compelled to point out that they’re actually entirely different words.
(PS a good example of the process I’m talking about is there/their/they’re or you’re/your – most people probably know exactly how to use these properly, but a sizeable minority are only ever just guessing. We get by – you see the mistakes so often that by now you probably don’t even bother pointing them out.)