The auction mechanism hasn’t received much attention, probably because the expectation is that it only comes into play in black swan events. But with roadmap items like unbonded minipools and shared validators, it might play a more central role in the future. And if we want to market the RPL collateral as insurance for rETH stakers, we ought to make sure that the mechanism is sound.
In my view, these should be the minimum requirements for the mechanism to be in line with the expectation that RPL serves as insurance:
If a minipool somehow ends up with a balance below 16 ETH on exit while there is x RPL staked on the node, the first priority should be that the rETH stakers are made whole in that 16 ETH is returned to them. Returning less than 16 ETH should only happen if x RPL is not enough to cover the shortfall, meaning that the Node Operator loses all of their staked RPL in the process. Any gas costs associated with the auction/liquidation, inefficiencies of the mechanism or sudden price slippage should be borne by the node operator and not by rETH stakers. Minimizing the cost to the node operator should be a secondary concern.
It’s my understanding that the current solution falls short of that. Only a portion of the RPL is reserved for auction, just enough to cover the loss in a best case scenario. The expected outcome is that RPL is sold for lower than market value, resulting in realized losses for rETH stakers, while the node operator gets to keep a portion of their RPL. In addition, the way that the auction is conducted leaves the door open for bidders to game the system by placing a partial bid and potentially extract extra value at the cost of rETH stakers.
Given that there are multiple open source solutions for the general problem of covering a debt with collateral in another token (thinking of Maker, Compound, Aave,…), that have been audited and battle tested with billions of dollars on the line, why does Rocket Pool need a unique solution and can’t fork an existing one? What is the rationale for the current design? In particular, what is the rationale behind partial bids getting a better price over time?