EigenLayer, Karak, and LRTs: The Restaking Risk Loop

Ethereum's staking system pays validators to secure one thing: Ethereum. Restaking is the observation that the same collateral, already locked and already subject to penalties, could simultaneously back other systems that need economic security and cannot easily bootstrap their own.

That is a genuinely useful idea. Oracles, bridges, data availability layers, and sequencers all need a large pool of capital that loses money if it misbehaves, and building one from scratch is expensive. Restaking rents that capital from a pool that already exists.

The price is that the collateral now has more ways to be taken. This article covers how much more, how the protocols bound it, what a liquid restaking token changes about the position, and why leveraging one produces a risk profile that has very little to do with the yield it advertises.

1. What restaking actually sells

The demand side of a restaking network is a set of services that want economic security. EigenLayer calls them Actively Validated Services. Karak's protocol specification calls them Distributed Secure Services, describing a DSS as a service that "coordinates the tasks and rewards for operators" and that has the authority to slash operator funds for inadequate performance.

The supply chain in both designs has the same three roles:

The whole economic proposition sits in that third step. Restaking yield is compensation for accepting additional slashing conditions written by third parties. Any evaluation that treats it as free extra yield on the same risk has misread the trade.

2. How EigenLayer bounds the slashing surface

The most common criticism of restaking is that one deposit ends up pledged to many services at once, so a single bad day could see the same ETH slashed repeatedly. EigenLayer's allocation design addresses this directly, and getting it right matters because the popular version of the risk is not the real one.

Magnitudes and unique stake

EigenLayer tracks allocation through magnitudes. Per the protocol's contract documentation, each operator has a max magnitude per strategy, described as "the maximum magnitude that can be allocated between all operator sets," starting at 1 WAD (1e18), representing 100% of allocatable stake. The currently allocated portion is the encumbered magnitude.

The constraint that produces unique stake is stated plainly: "An operator cannot allocate > 100%; therefore, a strategy's encumbered magnitude can never exceed that strategy's max magnitude." The documentation notes that this ensures stake is not double-pledged across multiple services.

An operator becomes slashable by a service only when three conditions hold together: the operator is registered for the operator set or remains slashable after deregistering, the service has added the strategy to that operator set, and the allocation has nonzero current magnitude.

When a slash occurs, it "reduces allocation.currentMagnitude by the slashed magnitude" and pushes a new entry reducing the operator's max magnitude. The reduction is permanent.

What this design actually tells you

The "same ETH slashed by everyone" fear is addressed by the architecture. A given unit of stake is allocated to one operator set at a time. A service can only take from what was explicitly pledged to it. This is a real safeguard and it should change how you weigh the risk.

The real exposure is fragmentation, not duplication. An operator can split their magnitude across several operator sets. If 30% is allocated to one service and 30% to another and both slash their full allocation, the operator loses 60%. Slashes compound up to but not beyond 100%. Your effective exposure is therefore the sum of your operator's allocations across services, and that is a number worth asking your operator for.

Delegation is the concentrated risk, not the protocol. Unique stake protects you from a protocol-level double-pledge. It does not protect you from an operator who allocates aggressively across many services with weak conditions. The operator's allocation policy is the single most important variable in a restaking position and it is chosen by them, not by you.

The exit is slower than the risk

This is the part that turns a manageable risk into a structural one.

EigenLayer's contract documentation specifies a mainnet allocation delay of 126,000 blocks, roughly 17.5 days, and a mainnet deallocation delay of 100,800 blocks, roughly 14 days. Karak's protocol specification sets a 9-day minimum withdrawal delay for stakers, explicitly "to prevent front-running slashing events," with operators facing a similar delay when unallocating from a service.

3. Karak and the alternative implementations

EigenLayer is not the only implementation of the shared security idea, and comparing designs is a useful way to see which choices are inherent and which are decisions.

Karak's protocol specification, published with its 2024 audit, describes a design with a distinct approach to bounding slashing. Rather than allocating proportional magnitude, "DSSs must set a maximum slashable percentage beforehand," and a slashing committee can veto unfair slashing events.

The trade-off in the veto mechanism is worth naming. A committee that can block an unjust slash also introduces a trusted party into a system whose value proposition is trust minimization. Neither choice is obviously correct. What matters for an allocator is knowing which one you are relying on.

4. What a liquid restaking token adds and removes

A restaked position has an obvious usability problem: the capital is locked behind a multi-week exit and cannot be used for anything else. A liquid restaking token solves it the same way liquid staking tokens solved the equivalent problem for validators. The protocol takes deposits, runs the restaking position, and issues a transferable token representing a claim on it.

What it adds

Immediate transferability. The 14-day exit is replaced by a sale on a decentralized exchange that settles in one block.

Composability. The token becomes usable as collateral, as a liquidity pool asset, and as the underlying for yield tokenization on platforms like Pendle.

Outsourced operator selection. The LRT protocol picks operators and allocation policy, which is a service if they are good at it.

What it removes

A direct claim on the underlying at par. You now hold a token whose value depends on both the restaked position and the LRT protocol's own solvency and honesty.

Visibility into the slashing surface. The operator set, the services backed, and the allocation across them are chosen by the LRT protocol. Your actual slashing exposure is whatever their policy produces.

A meaningful share of the yield. Fees stack at every layer: the LRT protocol takes a cut of rewards, operators take a commission, and the restaking protocol may take its own. Each layer takes a percentage of the rewards while the slashing conditions apply to the full principal.

The exit liquidity substitution

The most important consequence is subtle. An LRT does not make the underlying position liquid. It gives you a different exit: selling the token to another buyer instead of redeeming it from the protocol.

In calm markets those two are nearly equivalent, because arbitrageurs keep the token near redemption value by buying any discount and redeeming it. In stress they diverge sharply, because the redemption leg of that arbitrage is subject to the same multi-week queue as everyone else's. The mechanism that normally holds the price at par is switched off for exactly as long as the delay lasts.

So the practical exit price of an LRT under stress is set by secondary market depth, not by the value of the underlying. That single fact drives everything in the next section.

5. The leverage loop

Because LRTs are transferable, they can be used as collateral. And because an LRT's yield is typically above the cost of borrowing ETH, the arbitrage is obvious and heavily used.

The mechanics

Deposit LRT as collateral, borrow ETH against it, convert the ETH into more LRT, redeposit, and repeat. At an 80% loan-to-value ratio, total exposure converges toward five times equity if you loop indefinitely, and six practical rounds get you to roughly four times.

Starting with 100 units of equity, six rounds hold about 395 units of LRT collateral against 295 units of ETH debt.

The return

Assume the LRT yields 4.2% and ETH borrowing costs 3%:

• Income: 4.2% of 395 = 16.6 units
• Interest: 3.0% of 295 = 8.9 units
• Net: 7.7 units on 100 of equity, or about 7.7%

A 1.2 point spread became 7.7%. That is the pitch, and the arithmetic is right.

The risk

Reading Figure 1

The two lines are in the same units — percent of equity — so they can be compared directly, and the comparison is the point.

Start at 80% and read across. The green line sits near 7.7% and the red near 12%. That is the position described above: it pays about 7.7% a year and it is gone if the LRT loses about 12% against ETH.

Now drag the loan-to-value up. Income rises slowly, because each extra round adds a little more of the same 1.2 point spread. The survivable fall collapses, because each round adds debt against the same equity. The asymmetry is the argument: you are not buying more yield so much as selling more insurance.

Then hold the red line against the two paragraphs below. A 5% slash is roughly 20% of equity at this leverage, and the redemption queue that would normally defend the LRT's price is unavailable for about two weeks in exactly the conditions that threaten the position. The red line assumes you can be liquidated at a market price; the callout below explains why that price can fall much further than the underlying is worth.

So read the red line first, and read it as a secondary-market number. The right question is not whether the LRT is worth par. It is how far below par it can trade while redemption is queued — and Figure 1 says how much of that you can survive.

Using the standard health factor definition with an 85% liquidation threshold:

Health Factor = (Collateral Value × Liquidation Threshold) ÷ Total Borrow Value

At par that is (395 × 0.85) ÷ 295 = 1.14. Now solve for the LRT-to-ETH exchange rate at which the health factor reaches 1:

395 × x × 0.85 = 295, so x = 0.879

A 12.1% decline in the LRT's exchange rate against ETH liquidates the position.

What this example actually tells you

The payoff is a sold option, not a yield. You collect roughly 7.7% per year and lose a large share of equity if the exchange rate falls about 12%. That is economically the same shape as selling an out-of-the-money put on the LRT peg, with the premium paid as yield. It is a legitimate trade and it must be sized like a short option position rather than like a deposit.

A 5% slash costs 20% of equity. Slashing reduces the value of the underlying restaked position, which flows through to the LRT's exchange rate. At 4x exposure, a 5% haircut on 395 units of collateral is about 20 units against 100 of equity. The loop multiplies slashing exposure by exactly the same factor as it multiplies yield, but from a much smaller base.

And here is the part that makes it worse than the arithmetic suggests. Your liquidation trigger is a secondary market price. The mechanism that defends that price is redemption arbitrage. Redemption takes roughly two weeks. So in the exact scenario where the loop is at risk, the price defense is unavailable for longer than the liquidation takes to execute.

The correct sizing question is therefore not "will this asset lose value" but "how far can this asset's secondary market price fall while redemption is queued." Those are different numbers, and the second one is much larger.

6. Why these risks do not diversify

Holding several different LRTs feels like diversification. Frequently it is not, because the exposures share their dependencies.

The practical test is simple: write down the operator set, the services, the lending market, and the exit pool for each position you hold. If the lists overlap, you have one position in several wrappers, not several positions.

7. An evaluation checklist

8. Conclusion

Restaking is a real financial innovation with a clearly stated price. Collateral that was already securing Ethereum takes on additional slashing conditions written by third parties, and is paid for accepting them. Protocols have done meaningful work to bound that surface. EigenLayer's unique stake design prevents the same stake being pledged to multiple services, and Karak's specification bounds each service to a declared maximum slashable percentage with a committee able to veto unjust slashing.

What no design removes is the exit delay, and the exit delay is where the risk actually concentrates. A roughly two-week withdrawal window is correct protocol engineering, since instant exits would let participants escape penalties they have earned. It also means a restaking position cannot be stopped out, and it means the arbitrage that normally holds a liquid restaking token near its underlying value is unavailable during exactly the periods when the token is under pressure.

That combination is what makes leveraged restaking loops so much riskier than their spread suggests. The loop pays a few points of net yield and liquidates on a roughly 12% move in an exchange rate whose defense mechanism is structurally too slow to arrive. Sized as a short option position with a wide strike, it is a defensible trade. Sized as a yield product, it is the single most reliable way to lose a large amount of capital in DeFi while holding an asset that was, in the end, worth par.

For how restaking compares with other advanced on-chain strategies, see advanced DeFi yield and hedging strategies. For how the lending markets these loops run through actually price and liquidate collateral, see leveraged yield and credit accounts and soft liquidations and Curve LLAMMA.

Frequently asked questions

Sources and further reading

Primary protocol documentation:

• EigenLayer contracts — AllocationManager documentation
• Karak restaking — protocol specification
• Aave — Health factor and liquidations

Related CoinBeaver articles:

• Advanced DeFi yield and hedging strategies
• Leveraged yield and credit accounts: Gearbox vs Morpho Blue
• Soft liquidations: how Curve LLAMMA works
• How Pendle yield trading works
• DeFi lending explained
• Crypto liquidations explained

This article is educational and is not financial advice. All yields, loan-to-value ratios, liquidation thresholds, and position sizes used above are illustrative examples rather than live protocol parameters. Restaking protocols, operator sets, supported services, slashing conditions, fee structures, and withdrawal delays change over time, and the Karak figures cited come from a 2024 specification that may not reflect the protocol's current state. Verify every parameter in the relevant protocol's own documentation before committing capital.