This 'Frax dance' is always happening and uses AMM game theory to test different ratios of collateralization, incentivize recollateralizing through arbitrage swaps, and redistribute excess value back to FXS holders through a buyback swap. The protocol starts at a 100% collateral ratio at genesis and might or might not ever get to purely algorithmic. The novel insight is to use market forces itself to see how much of a stablecoin can be algorithmically stabilized with its own seigniorage token so that it keeps a tight band around $1 like fiatcoins. Purely algorithmic/rebase designs like Basis, ESD, and Seigniorage Shares have wildly fluctuating prices as much as +-40% around $1 that take days/weeks to stabilize before going through another cycle. This is counterproductive and assumes the market actually wants/needs a stablecoin with 0% collateralization. Frax doesn't make this assumption. Instead, it measures the market's preference and finds the actual collateral ratio which holds a stablecoin tightly around $1, periodically testing small differences in the ratio when the price of FRAX slightly rises/drops. Frax uses AMM concepts to make a real-time fractional-algorithmic stablecoin that is as fast at price recovery as Uniswap is at keeping trading pools correctly priced.
As this system gets more efficient and the velocity of the system increases, collateral pools can include other assets instead of stablecoins like volatile crypto such as ETH and wrapped BTC. As the price of the volatile asset rises, users will use the buy back shares function to distribute the excess value to FXS holders. When the price of the volatile collateral drops, there is an instant arbitrage opportunity to put in more crypto for discounted FXS to keep the collateral ratio at the target. Just like a Uniswap trading pair keeps its constant product function balanced, the Frax Protocol keeps its target collateral ratio balanced to what the market needs for FRAX to be $1.