If you take into account the fact that you can obtain the order-of-magnitude for most "human" numbers without actually focusing on them, then putting the most-significant digit first in the stream makes sense. 30 is easy to see, even if you look at the 3 you can at least tell there's two digits. 12,048 is pretty easy. Even 8,293,254 isn't that hard to order-of-magnitude from the near-fovea while you're focused on the 8. By the time you get into the larger numbers that are hard, well, they're hard under either order so it's a wash.
At reading, one may be indeed hit at first by the most significant digit ('8' in your last example) that is somewhat useful, but the reader can't easily figure out digit's order of magnitude without reading the entire number. There's actually more. The reader can't spell the number right away, as it has to be read entirely and compiled in mind first. This compilation involves counting all digits in the given number to determine the number's size, then starting again by interpreting and spelling accordingly in its order each digit and its designation. This double-pass doesn't appear that efficient to me. On the other hand, if the number from your last example would be written backwards (452'392'8), its reading and interpretation would come efficiently in one pass ("four fifty and two hundred, three ninety and two hundred thousand, and eight million"). Even when all we're interested in would be the most significant digit and its order of magnitude, that would still be faster to determine (i.e. again, in one pass).
"but the reader can't easily figure out digit's order of magnitude without reading the entire number."
Yes, they can; that was my entire point. Human-sized numbers fit within the eye's fovea. Numbers that don't fit easily within the fovea require effort under either scheme. Your model of having to go digit-by-digit is a computer's scanning model, not a human perception model; we do not scan that way, we scan in fovea-sized chunks, which are smaller than people may think because the brain is very good at interpolating before the information hits the conscious mind, but is still large enough to fit "millions" in quite comfortably.
