So it sounds like the incentive model is kind of fundamentally broken. Is it correct to say that miners are incentivized to A) store trash data in a way that lets them mine with zero actual storage backing it up and B) consolidate their storage into the same hard drive, until there exists only a single copy of the data on the entire network?
Yes. Their counter argument to (A) that the trash data will cost more to store than the user gets for retrieving it (though this isn't true if the attack is successful and results in them being the only miner/retrieval server), and (B) they don't seem to consider a problem.
There are other issues— e.g. arbitrary data can be censored and the only harm is lost mining revenue proportional to the censored data. (e.g. you could censor 0.1% of all data, and probably make every file unreadable (just by censoring the first word of it) but only be unable to answer 0.1% of the file queries).
I don't really see the point— having a file storage service that you can pay for with cryptocurrency make sense, but integrating it in this manner seems to serve no technical purpose other than to fuel investment speculation and otherwise just makes the system worse.
