The arbitrage loop that holds a peg in place
Across all three designs, day-to-day peg maintenance relies on the same basic arbitrage logic, even though what backs the redemption differs completely. If a stablecoin trades below its one-dollar target on the open market, an arbitrageur can buy it cheaply and redeem it through the protocol's official channel for a full dollar of value (or a full dollar's worth of underlying collateral), pocketing the difference. That buying pressure on the open market, plus the reduction in circulating supply from redemption, pushes the price back up toward the peg. Symmetrically, if the stablecoin trades above one dollar, someone can mint new tokens at the official rate and sell them on the open market above that rate, again pocketing the spread, which increases supply and pushes the price back down. Both directions are profit-seeking behavior that happens to have the side effect of defending the peg.
The crucial word in that description is 'official channel'. This entire loop only functions if minting and redemption actually work as advertised, reliably, at the rate they're supposed to, without excessive delay or restriction. For a fiat-collateralized token, that means the custodian must actually honor redemptions promptly. For a crypto-collateralized token, that means the on-chain mint and redeem or liquidation processes must function as coded, with oracles reporting real prices. For an algorithmic token, there often isn't a hard redemption guarantee at all, only an incentive to participate, which is precisely why its arbitrage loop is the least dependable of the three under stress. Whenever redemption becomes unreliable, whether through custodian delay, contract congestion, or collapsing incentive participation, the arbitrage loop weakens exactly when it's needed most, and the market price can drift away from the peg with nothing pulling it back.