Thanks @alexcelo for replying in detailled.
Option 1) would still be incompatible with Granda Mento focusing on USD → Celo → cUSD as per the original discussion of Granda Mento: Discussion on Granda Mento: Enabling Larger Stablecoin Mints - #40 by trevor
In theory, I think you could do option 2 (although it seems complicated for foundations) but again I don’t think would be in the spirit of Discussion on Granda Mento: Enabling Larger Stablecoin Mints - #40 by trevor .
I also echo @thezviad 's point that the multisig should be a simple check, which is doable when it’s USD → CELO → cUSD. Yes, this does require the proposer to share evidence of the fiatUSD → Celo transaction and aggregate price.
I do think there is a good - separate to Granda Mento - discussion to be had on how to reduce Mento slippage because of using x*y=constant (e.g. move to a Uniswap v3 approach or an order book type DEx). There’s also a related discussion on how to size the portion of the reserve that is used for exchanges (a tricky question that involves reserve security too).