Summary
LlamaRisk presents this insight to the Aave DAO and community regarding Ethena’s governance proposal to adopt a dynamic sUSDe unstaking cooldown mechanism. As a member of the Ethena Risk Committee and a risk service provider to Aave, we are uniquely positioned to assess the implications of this change for both Ethena’s protocol stability and sUSDe’s role as collateral within Aave. This report draws on our independent onchain analysis and the framework co-developed with Blockworks Research and Kairos Research. The evidence demonstrates that a dynamic cooldown, supervised by the risk committee with clearly defined escalation criteria, is a sound improvement over the current fixed 7-day period. Beyond improving user experience, the change is expected to reinforce sUSDe secondary market price stability, directly relevant to sUSDe’s soundness as collateral on Aave.
Ethena’s Liquid Buffer
Ethena holds backing assets across onchain and custody wallets on Ethereum mainnet and several L2 networks (Base, Mantle, Plasma). These assets serve as the liquid buffer that absorbs redemption demand when users unstake sUSDe or redeem USDe directly through the EthenaMinting contract. As of March 12 2026, the total on-chain observable buffer stands at approximately $4.14B against a USDe supply of 5.9B with 3.51B staked into sUSDe.
Blockworks classified these assets into three tiers based on settlement speed. Tier 1 assets (USDC, USDT, USDtb, PYUSD) can be settled within one day and represent the most immediately available liquidity. Tier 2 assets (Aave aTokens such as aUSDC, aUSDT, aUSDtb, aPYUSD, along with sDAI and sUSDS) are constrained within the utilization and may require up to two days to withdraw from lending protocols. Tier 3 assets (Morpho vaults) are subject to even larger utilization constraints and may require up to five days to fully exit.
Source: LlamaRisk, March 13, 2026
The following table summarises the current on-chain holdings across all tracked chains.
Ethereum Mainnet
| Token | Balance | Tier | Settlement |
|---|---|---|---|
| USDC | $1,495.3M | 1 | 1-day |
| PYUSD | $1,126.6M | 1 | 1-day |
| USDtb | $517.3M | 1/2 | 1% instant / 99% banking hours |
| aUSDtb | $196.2M | 2 | 2-day |
| USDT | $152.5M | 1 | 1-day |
| aPYUSD | $103.1M | 2 | 2-day |
| aUSDT | $90.3M | 2 | 2-day |
| SENTORAPYUSDCOREV2 | $49.9M | 3 | 5-day |
| STEAKUSDTBETHENA | $49.2M | 3 | 5-day |
| aUSDC, sUSDS, STEAKUSDCETHENA | < $1M each | 2/3 | 2–5 day |
| Ethereum subtotal | $3,780.3M |
L2 Deployments
| Chain | Token | Balance | Tier |
|---|---|---|---|
| Plasma | APLAUSDT0 | $645.5M | 2 |
| Mantle | aManUSDT0 | $248.1M | 2 |
| Base | STEAKPRIMEUSDC | $213.7M | 3 |
| L2 subtotal | $1,107.4M |
Tier-Level Summary
| Tier | Settlement | Current Balance | Share |
|---|---|---|---|
| Tier 1 | 1-day | $2.78B | 57% |
| Tier 2 | 2-day | $1.80B | 37% |
| Tier 3 | 5-day | $0.31B | 6% |
| Total | $4.89B |
USDtb occupies a special position in this classification. While it is a stablecoin, its redemption depends on banking hours for the underlying T-bill settlement. Accordingly, 1% of USDtb holdings are classified as Tier 1 (representing the 24/7 instant buffer maintained by the issuer) and the remaining 99% as Tier 2.
It is important to note that the on-chain buffer represents only the observable portion of Ethena’s liquidity. Additional reserves are held within custodian accounts (Coinbase, Anchorage, Kraken), generally in USDT form, which are not captured by on-chain balance queries. The figures presented here therefore represent a conservative lower bound.
Buffer Resilience Under Pressure
The liquid buffer must absorb two types of outflow: sUSDe unstaking (users exiting the StakedUSDeV2 vault) and direct USDe redemptions through the EthenaMinting contract. Over the observation period, the protocol processed $10.3B in unstaking volume across 22,456 events and $14.6B in direct redemptions across 11,702 events. Combined, these represent the total demand-side pressure on Ethena’s reserves.
Queue pressure measures the intensity of outflow relative to recent history. It is defined as the ratio of daily combined outflow to the 14-day rolling average.
QP(t)=avg of 14 non-overlapping 24h windows before [t−24h]∑(unstaking+redemptions) in [t−24h,t]
A queue pressure of 0.5-1.0x indicates normal activity (unstaking and redemption pressure mild or in a declining phase); values above 2.0x indicate outflow is running at double the recent baseline, and values above 5.0x represent extreme tail events. Over the full period, median queue pressure was 0.51x (indicating that most days see below-average outflow). The 95th percentile reached 4.01x, the 99th percentile 8.86x, and the historical maximum 13.49x. Days with queue pressure exceeding 2.0x occurred 11.8% of the time (72 days), and days exceeding 5.0x occurred 3.3% of the time (20 days).
Source: LlamaRisk, March 13, 2026
The single largest daily outflow was $1,705M on 11 October 2025.
Source: LlamaRisk, March 13, 2026
Coverage adequacy ratios evaluate whether the buffer can absorb historically severe redemption demand. Blockworks derived P99 benchmarks by computing the 99th percentile of cumulative net USDe redemptions observed over rolling 1-day, 3-day, and 7-day windows across the protocol’s history. These benchmarks (4.1% of supply for the 1-day horizon, 9.2% for 3-day, and 12.0% for 7-day) represent near-worst-case outflow scenarios: only 1% of historical periods saw higher redemption pressure. Each coverage ratio is then calculated as the available liquidity at the corresponding tier divided by the product of current USDe supply and the P99 benchmark. For example, 1-day coverage equals Tier 1 balance / (supply x 4.1%). A ratio above 1.0x means the buffer exceeds the P99 demand level; We recommend maintaining at least a 1.5x safety threshold.
Over the full observation period, 1-day coverage averaged 2.80x with a median of 2.78x. The current value is 8.40x. 7-day coverage averaged 2.91x with a median of 3.32x and currently stands at 6.88x. The coverage ratios have trended sharply upward since mid-2025 as Ethena diversified its backing composition and grew the liquid buffer. The current values represent the most comfortable liquidity position in the protocol’s history.
Source: LlamaRisk, March 13, 2026
Secondary Market Efficiency and the 7-Day Cooldown
The current fixed 7-day cooldown creates a structural friction for users seeking to exit their sUSDe position. Rather than waiting the full cooldown period, users who need immediate liquidity sell sUSDe on secondary markets (primarily Curve and other DEX pools). This selling pressure periodically pushes the sUSDe market price below its fair value, creating a discount that arbitrageurs must bridge.
Fair value is computed as the sUSDe/USDe exchange rate multiplied by the USDe/USD price. Over the observation period of 607 daily observations, sUSDe traded at a mean deviation of -0.168% from fair value. The maximum discount reached -1.270% and the maximum premium +0.079%. Overall, sUSDe traded at a discount on 91.9% of observed days (558 days) and at a premium on only 8.1% (49 days). This persistent discount is a direct consequence of the 7-day cooldown: it represents the cost that impatient sellers pay to bypass the waiting period.
Source: LlamaRisk, March 13, 2026
A shorter cooldown narrows the arbitrage window. When the cooldown is 1 day instead of 7 days, arbitrageurs can buy discounted sUSDe, initiate unstaking, and receive USDe within 24 hours rather than a week. This compresses the maximum discount the market can expect, improving price efficiency for all holders. The 7-day cost of capital that currently anchors the discount floor would shrink to a 1-day cost, reducing the structural discount by roughly 6/7ths in equilibrium (even if we can still expect the heavy discount tails during stress events).
Ethena has not actively incentivized sUSDe DEX liquidity in recent months. As a result, the majority of exit volume flows through the direct unstaking path. A dynamic cooldown that reflects actual backing liquidity would align the unstaking friction with the protocol’s ability to process redemptions, rather than imposing a fixed delay regardless of conditions.
Technical Mechanics of sUSDe Unstaking
The sUSDe unstaking flow involves three contracts: StakedUSDeV2 (the main ERC4626 vault), USDeSilo (the escrow contract), and USDe (the underlying ERC20 stablecoin). The process follows a precise sequence of steps.
- The user calls
cooldownShares(shares)on the StakedUSDeV2 contract. - The sUSDe shares are burned immediately at the current exchange rate (
totalAssets / totalSupply). - The equivalent USDe amount, calculated via
previewRedeem(), is transferred from the vault to the USDeSilo escrow contract. - A cooldown timer is initiated:
cooldownEndis set toblock.timestamp + cooldownDuration. - After the cooldown period has elapsed, the user calls
unstake(receiver)to claim. - The USDeSilo transfers the locked USDe to the specified receiver address.
Several aspects of this design are relevant to the dynamic cooldown proposal. First, the user earns no yield during the cooldown period. Since the sUSDe shares are burned at step 2, the user’s capital sits idle as USDe in the Silo. The yield that would have accrued on those shares is instead distributed to remaining stakers, creating a modest yield forfeiture penalty for unstakers. Second, there is no time limit to claim after cooldown expiry; the user may call unstake at any point after cooldownEnd. Third, if a user calls cooldownShares again before claiming, the timer resets to a new cooldownEnd and the amounts accumulate, penalizing users that start new cooldowns before the previous amount is claimed.
Critically for the dynamic cooldown proposal, users who have already initiated unstaking when the global cooldownDuration changes are not affected. The cooldownEnd timestamp is fixed at initiation time. If the protocol shortens the cooldown from 7 days to 1 day, existing users in the queue still proceed with their original 7-day period. Conversely, if the cooldown is lengthened, only new initiations are subject to the longer duration.
The yield distribution mechanism provides an additional safeguard against arbitrage under shorter cooldowns. Ethena distributes rewards via transferInRewards() on the StakedUSDeV2 contract approximately every 8 hours. Each distribution triggers a linear vesting period of 8 hours (the VESTING_PERIOD constant), during which the exchange rate rises gradually rather than in a discrete step. On-chain analysis of the past year of RewardsReceived events confirms this cadence: 1,098 distributions were observed over 367 days, with a median gap of exactly 8.0 hours between consecutive events. The average distribution size was approximately 189,127 USDe, and the implied annualized yield from these distributions is approximately 5.8%.
Source: LlamaRisk, March 13, 2026
Source: LlamaRisk, March 13, 2026
Even with a 1-day cooldown, the 8-hour vesting eliminates the viability of arbitrage or front-running strategies: a user staking immediately before a distribution would need to wait at least 1 day, during which only approximately 3 distributions vest. The yield from those 3 distributions (roughly 1/365th of the annual yield, or approximately 0.016% at current rates) would naturally disallow the user from receiving distributions that would represent rewards held for a longer time than the user would have held the deposits for.
Implications for sUSDe as Aave collateral
sUSDe is a significant collateral asset on Aave, and the cooldown mechanism directly influences its secondary market behavior. As documented in this report, the fixed 7-day cooldown has produced a persistent discount on sUSDe (mean deviation of -0.168% from fair value, with a maximum discount of -1.270%), as impatient sellers bypass the waiting period via DEX pools. We could also argue similar effects for other sUSDe-based assets, namely PT-sUSDe tokens, which are priced with harsher temporary discounts as sUSDe experiences temporary de-pegs. A shorter, dynamically managed cooldown compresses this structural discount by reducing the cost of capital that arbitrageurs must bear, tightening secondary market pricing around fair value. For Aave, this translates to a reduced likelihood of iliquidity.
However, the dynamic nature of the cooldown introduces a new variable for Aave’s risk framework to monitor: a sudden extension of the cooldown period during stress could temporarily widen discounts before arbitrageurs adjust. LlamaRisk will continue to monitor cooldown changes and their impact on sUSDe market dynamics as part of our ongoing risk oversight for both Ethena and Aave.
Conclusion
The unstaking cooldown will be supervised by the risk committee, using the framework co-developed together with Blockworks Research and Kairos Research as the reference for decision-making. The committee will monitor backing composition, coverage adequacy ratios, and queue pressure to determine the appropriate cooldown level at any given time.
The most important principle guiding the committee’s oversight is predictability. The cooldown should not change in a volatile or erratic manner. Users and integrators need to have reasonable expectations about the unstaking timeline, and frequent oscillations would undermine confidence in the mechanism.
Short cooldowns are appropriate when the backing is predominantly composed of liquid stablecoins and coverage ratios are comfortably above the safety thresholds. Longer cooldowns are warranted when the portfolio shifts toward less liquid positions, when coverage ratios approach or fall below critical levels, or when queue pressure indicates sustained abnormal outflow demand. The dynamic mechanism ensures that the unstaking friction imposed on users is proportional to the actual conditions of the protocol rather than a fixed constant that ignores the state of the backing.
Disclaimer
This review was independently prepared by LlamaRisk, a DeFi risk service provider funded in part by the Aave DAO. LlamaRisk serves as a member of Ethena’s Risk Committee and an independent attester of Ethena’s PoR solution. LlamaRisk did not receive compensation from the protocol(s) or their affiliated entities for this work. The information should not be construed as legal, financial, tax, or professional advice.