A stablecoin is designed to trade at a constant value against a reference currency, and most of the time it does. The mechanism holding it there is narrower than it looks, and it can fail in a small number of identifiable ways.
The peg is enforced by arbitrage
Nothing in the software forces a stablecoin to trade at its reference value. The price on exchanges is set by buyers and sellers exactly as it is for any other traded asset.
What holds it steady is the ability to redeem. If the token trades below its reference value, a trader can buy it cheaply, redeem it with the issuer for full value, and pocket the difference.
That trade pushes the price back up, and the reverse trade pushes it down when the token trades above value. The peg is therefore a consequence of arbitrage, and it lasts only while arbitrage remains practical.
Reserve doubt breaks the arbitrage first
The redemption trade only makes sense if the trader believes the issuer can pay. That belief rests on the reserves held behind the tokens in circulation.
When holders begin to question what those reserves contain, or whether they can be converted quickly, the arbitrage stops looking risk-free. Buying a discounted token becomes a bet on the issuer rather than a spread trade.
At that point the correcting force weakens exactly when it is needed most. The discount widens because the people who would normally close it have stepped back.
Redemption friction matters as much as reserves
Many issuers restrict redemption to vetted institutional counterparties, apply minimum sizes, or process requests over a settlement cycle rather than instantly. Those restrictions are ordinary risk controls in calm conditions.
Under stress they become the constraint. If redemption takes days, the arbitrage trader carries the price risk for that whole period, and will only accept a much larger discount to do so.
A peg can therefore wobble while reserves remain entirely sound, simply because the route back to the reference currency is slow or narrow at the moment it is tested.
Collateralised designs face a different failure
Some stablecoins are backed not by cash-like reserves but by other volatile assets locked in overcollateralised positions. The excess collateral is the buffer that keeps the token fully backed.
If the collateral falls in value quickly, automated systems must sell it to restore the buffer, and those sales land in a market that is already falling. The mechanism designed to protect the peg pushes prices further down.
Designs that rely on a second token to absorb losses face a sharper version of the same problem, because that token typically falls hardest exactly when it is being issued to cover a shortfall.