Okay, what I'm reading into this is that in order to borrow 15 coins (sell 15 repocoins worth of bitcoin with an offer to buy the same amount of bitcoin back for ~15.005 repocoins the following day) the borrower must "lock" collateral worth a bit more than 15 repocoins for the duration of the time he's borrowing.
If this collateral is repocoins, then obviously his borrowing makes no sense because repocoins are fungible. whatever purpose he has for the repocoins he intended to borrow would be just as well served by the repocoins he's locking up for the duration. Therefore I conclude that the collateral here must be in some other asset. But if the collateral is in Bitcoins, then he's locking up the same amount of Bitcoins that he's loaning out, so wouldn't he be charging additional interest because this is tying up twice as much of his assets as are actually appearing on the market?
Over-collateralization is standard in repo exchanges. The collateral here is mostly bitcoins, plus the small amount of repocoins required to pay the interest.
Your point about the missing interest on the bitcoins is correct though. Just as the buyer is lending the seller repocoins, the seller is effectively lending the buyer bitcoins, and the buyer might legitimately like interest on this, if they didn't expect compensating increases in the purchasing power of bitcoins. The Hotelling model of resource extraction would predict precisely such increasing bitcoin value, which was what I had in the back of my mind. However, it is an undesirable feature of what I presented previously that some of the medium run departures from Hotelling rule growth of repocoins will potentially be reflected in the value of bitcoins. I'll see if this can be remedied.