this was my exact thoughts at the end of 2017 when Maker released their first version. Why instead of CDPs just don't use one big liquidy pool which initially funded with sufficient resources? If it is big enough - it would initially overcollateralized the system (when the cap is small) and can be used as a stabilization reserve, i.e. everyone would able to redeem their $1 stablecoin for $1 of ETH at any moment of time. I even try to design such a system https://stableunit.org/StableUnit-whitepaper.pdf
The main assumption that stablecoin with collateral/reserve has two values:
1) "redeemability value", i.e. the token without demand is worth $1 because you can redeem it for $1 of another asset with demand such as ETH.
2) network/native value: more people use the coin, higher the value. In the same manner as Dogecoin worth anything at all, like Metcalfe's law but for cryptocurrency. With an empirical estimation of growth asymptotic higher than NlogN.
The important part here that as the system grows bigger 1) grows linear but 2) grows to faster than linear which means there are some sizes of then system when 2) > 1). This practically means that you can back part of stablecoin with less liquid assets because they won't be used often (if all). It's one way to explain why USDT can function even with partially collateralized.
The problem with all of this, that one thing is to have system design (I'm convinced that there are many different decentralized stablecoin systems that can work) and another to build a working project with user adoption. I've spoken with many people at MakerDAO and they all know about the limitations/imperfection of the system. For oracles, they chose that expensive design because it was an "engineering approach which just works security". And as we see right now at stablecoins market cap, they were right to do so.
The main assumption that stablecoin with collateral/reserve has two values:
1) "redeemability value", i.e. the token without demand is worth $1 because you can redeem it for $1 of another asset with demand such as ETH.
2) network/native value: more people use the coin, higher the value. In the same manner as Dogecoin worth anything at all, like Metcalfe's law but for cryptocurrency. With an empirical estimation of growth asymptotic higher than NlogN.
The important part here that as the system grows bigger 1) grows linear but 2) grows to faster than linear which means there are some sizes of then system when 2) > 1). This practically means that you can back part of stablecoin with less liquid assets because they won't be used often (if all). It's one way to explain why USDT can function even with partially collateralized.
The problem with all of this, that one thing is to have system design (I'm convinced that there are many different decentralized stablecoin systems that can work) and another to build a working project with user adoption. I've spoken with many people at MakerDAO and they all know about the limitations/imperfection of the system. For oracles, they chose that expensive design because it was an "engineering approach which just works security". And as we see right now at stablecoins market cap, they were right to do so.