Liquid is doing a lot of work in the phrase liquid staking. I keep running into portfolios where someone holds a large slug of stETH or a similar token and mentally prices it as staked ETH with a sell button attached. In calm markets that is a fair description. In the exact moment you would most want the sell button, the price behind it can move away from you fast, and the gap between those two states is where people get hurt.
So it is worth being precise about what you actually own. A liquid staking token is a claim on staked ETH plus accumulated rewards, and that claim has two exits. You can redeem through the protocol, which pays full value but takes time. Or you can sell on the secondary market, which is instant but pays whatever the pool will bear. The two prices agree most of the time. Under stress they diverge, and the size of the divergence tells you exactly how expensive patience has become.
The two exit doors
Start with the protocol exit, since it defines what the token is theoretically worth. When you request a withdrawal from a protocol like Lido, the request enters a queue. If the protocol has enough ETH sitting in its buffer, from fresh deposits and rewards it has not yet restaked, the request gets filled quickly, often within a few days. If the buffer cannot cover it, the protocol has to exit validators from the beacon chain, and that is where the real clock starts.
Ethereum rate-limits validator exits. Only a certain number of validators can leave per epoch, a churn limit that scales with the size of the validator set, and it exists precisely for the scenario where a huge share of stake tries to leave at once. Under normal conditions the limit is invisible and withdrawals finalize in days. But if a meaningful fraction of all staked ETH heads for the door simultaneously, the queue stretches to weeks or months. The protocol exit is a promise to pay you in full at an unknown future date, and the date moves further away exactly when more people want out. That inverse relationship is written into the consensus layer, so no protocol can engineer around it.
The secondary market is the other door. The big Curve pool has historically been the main venue for stETH, plus assorted DEX pools and centralized exchange books for the larger tokens. This exit is instant, but you are selling your claim to someone who now has to hold it through the queue, and they will not do that for free. The price you get is redemption value minus a discount, and the discount is roughly the fee the market charges for waiting in your place.
Different tokens also wire the slow door differently, which matters more than the branding suggests. Rocket Pool's rETH redeems against a deposit pool that can run empty when exit demand is high, in which case you wait for it to refill or take the market price. Centralized tokens route redemptions through the issuer's own process and timeline. Some protocols have emergency modes that deliberately slow withdrawals during mass slashing events, which are precisely the moments when panic exits cluster. Reading the redemption docs for the specific token you hold takes maybe twenty minutes, and it is a better use of time than most of what passes for research in this market.
Why the discount widens exactly when you need the exit
Since withdrawals went live on Ethereum in 2023, there has been a real arbitrage anchoring LST prices. If stETH trades below redemption value, anyone can buy it, queue a withdrawal, and collect the difference when it finalizes. That trade is why discounts stay tiny in normal conditions, typically a few basis points, occasionally a few tenths of a percent when something spooks the market.
But look at what the arbitrageur is actually signing up for. They lock capital for the length of the queue, and they eat every risk that shows up during the window: a slashing event, a protocol bug, a further leg down in ETH that is expensive to hedge, the possibility that the queue itself keeps growing. The discount they demand is compensation for all of that, which makes the discount a function of queue length, and queue length a function of how many people are exiting. Mass exit lengthens the queue, a long queue makes the arb expensive, an expensive arb lets the discount widen. The mechanism that holds the peg in calm weather loosens its grip in a storm, by design.
The historical version is worth remembering even though the mechanics have improved since. In mid 2022, before withdrawals existed at all, stETH slipped to a discount of roughly 5 to 7 percent at the worst of it. There was no redemption arb back then, so the price was set entirely by sentiment and forced selling, much of it from leveraged funds unwinding. Withdrawals made the system healthier, but the shape of the failure carries over. The discount tends to gap rather than widen politely, arriving all at once when large holders are forced through the fast door together.
Pool composition as an early warning
Here is the part I actually find useful, because it is observable before the discount blows out. The main stETH pool on Curve is designed to sit near balance, roughly half ETH and half stETH, when flows are two-sided. Every seller who takes the fast exit pushes stETH into the pool and pulls ETH out. A balanced pool means exits are being absorbed by fresh demand. A pool drifting toward 60, then 70 percent stETH means sellers have been persistently outnumbering buyers, and the ETH side, which is the only side that matters for your exit, is draining.
The useful property is that composition deteriorates before price does. Arbitrageurs and liquidity providers defend the peg until their inventory runs thin, so the quoted discount can sit near zero while the depth behind it quietly disappears. In 2022 the pool skew worsened for weeks before the discount hit its widest point, and anyone watching the ratio instead of the price had a real head start.
So the read is straightforward. Track the pool ratio and its trend alongside the spot discount. Check how much ETH-side depth actually exists across the venues you would realistically use, meaning how much you could sell within, say, 1 percent slippage. And look at the protocol queue directly, how much is pending, how large the buffer is, how long recent requests took to finalize. Queue length rising while pool skew worsens is about the clearest exit-pressure signal this corner of the market gives you.
Sizing against the slow door
All of which leads to the sizing rule I actually use. Treat the fast exit as something that gets repriced in the exact scenario where you reach for it, and treat the slow exit as your real exit.
- Decide up front which door each part of the position uses. Whatever you might need on short notice has to fit through secondary liquidity at tolerable slippage, measured on a bad day, so assume depth is half of what you see quoted.
- Everything beyond that is queue money. Assume the queue takes weeks in the scenario where you are using it, because the scenario where you are using it is the scenario where everyone else is using it too.
- If you are looping, borrowing against the LST to buy more of it, run the liquidation math at a discount, and at a wider one than feels reasonable. The classic blowup is this exact loop. The discount widens, collateral marks down, liquidations force selling into the same thin pool, and the selling widens the discount further. Leveraged LST positions have a habit of becoming everyone's problem at once.
- Recheck pool composition and queue stats on a schedule rather than when they show up on your timeline. By the time an LST discount is trending on social media, the cheap exits are gone.
I still hold liquid staking positions, the yield is real and the underlying design is genuinely good. I just hold them with the discipline you would apply to any position whose liquidity depends on how many other people are reaching for it at the same moment. The pool ratios and queue stats live on a dashboard I glance at weekly, it is one of the things I keep on a Blockcircle scorecard, and the whole habit costs a few minutes a week. Cheap insurance for a position most holders never stress test until the market runs the test for them.