This is, at this moment, just an idea, there is a small amount of manpower behind it but it's on backburner - as I said, our "back of the envelope calculation" tend to indicate that it is not profitable. The only way to "balance the equation" is to bring in a "social benefit" - the benefit it will bring to the whole Steem community.
I actually thought further in detail - yes, I think the data should be striped, as you say, a kind of "virtual RAID" is what I had in mind. Then I envisage different quality tiers - from the lowest (for the really old pictures from long abandoned blogs) to the highest (for ... I don't know what yet, it will depend on the quality of the implementation). At the lowest quality the data would be only copied 3 or four times but the customer of the service would be able to pay more and indicate that it needs a higher assurance for his/her PEPE meme
Anyway, the reason I published is to attract people like you who are interested in contributing (even if only by challenging the design, if not with actual code or testing or something else)
So, looking forward to collaborating on this (when my team has a bit more time and we can launch this in collaboration with utopian probably)