We want to assure USD holders that they'll all eventually be able to sell if the price reaches some level and remains steady, but we also want to assure XTS holders that they won't be subjected to unbounded dilution over any finite time-frame. Here are my thoughts on how to accomplish that.
Back in March I proposed a compromise between options (1) and (3). Basically my idea was to use the amount of recently paid fees to create an upper bound on XTS dilution [1].
Thinking on the problem some more, I'm not sure if fees alone would provide adequate capitalization for the reserve. I suggest making the reserve increase each block, up to some capitalization limit. Then the reserve uses that money to close out undercapitalized shorts (using some deterministic algorithm based on the state of the market and ledger, subject to circuit-breaker type restrictions). Basically XTS holders will be charged interest (through dilution) in the beginning when the reserve is being funded, then the interest goes away when the reserve's capitalization limit is reached, then the interest is charged again to recapitalize the reserve after a black swan.
The cap limit could be a simple fraction of the reserve, but I like the idea of having it based on the maximum amount of XTS that would be needed to cover (say) a 2.5x rise in price.
If you want to charge a fixed rate of interest, you could just have the reserve capitalization increase each block by some fixed fraction of the total XTS supply. Or you could have the reserve decay exponentially to the capitalization limit to charge a higher interest rate to XTS holders when the reserve is badly undercapitalized, gradually decreasing the interest rate as the reserve nears its capitalization limit.
[1]
https://github.com/drltc/xts-proposal#insured-shorts