Worked example · 2022-06

Lido stETH depeg and Aave / Curve looped-leverage liquidation cascade — Ethereum L1 — 2022-05 to 2022-06

Loss: approximately several hundred million dollars in cumulative looped-leverage liquidations and forced-unwind realised losses across the Aave / Compound / Curve cohort during the May–June 2022 stETH depeg episode. Per Nansen's June 2022 forensic write-up, the stETH/ETH price ratio fell from ~1.00 to a low of ~0.931 on 2022-05-18, with persistent secondary-market discount through June [nansenstethdepeg2022]. The Curve stETH/ETH pool TVL contracted by more than 50% — from ~$4.08B on 2022-05-09 to ~$1.91B on 2022-05-12 — driven primarily by Three Arrows Capital (3AC) withdrawing ~$400M (~128k stETH + ~73k ETH) from the pool in a single transaction on 2022-05-12, and by Celsius withdrawing comparable amounts in the same window, leaving the pool's depth-at-price-tolerance materially reduced for the remainder of the cascade [nansenstethdepeg2022]. Academic forensic work (Heimbach et al., 2023) catalogued 442 leverage staking positions across Aave / Compound during this period totalling ~537,123 ETH (~$877M) of leveraged exposure, with cumulative repayments in the deleveraging cycle of ~136,069 ETH [heimbachleveragestaking2023]. Specific institutional losses are well-documented at the framing level: Celsius narrowly avoided full liquidation of its Aave stETH position (one of the largest individual stETH collateral positions on the protocol) by paying down its loan and withdrawing the collateral entirely [fintechcollectivesteth2022]; 3AC's stETH-related leverage was a contributing factor to its insolvency in late June 2022 [nansenstethdepeg2022]. Headline framing: $56M+ realised liquidations across smaller leveraged positions, with hundreds of millions in forced-unwind and de-risk costs across the institutional cohort. The case is the canonical pre-2024 LST/LSD depeg-and-cascade event and a load-bearing v0.1 datapoint for T14.003 sub-case (b) treated broadly to include LSTs, not only LRTs.
Recovery: none for the realised liquidations on the on-chain layer — the cascades cleared and the affected leveraged positions absorbed their realised losses. stETH price recovered toward NAV gradually through 2022-Q3 and Q4 as the Ethereum Merge approached and as confidence returned (full peg restoration tied to the post-Merge enabling of withdrawals through the Shapella upgrade in April 2023). The Celsius bankruptcy filing on 2022-07-13 ended its operations as a going concern; 3AC entered liquidation in late June 2022. The structural recovery primitive that mattered — restoring the 1:1 redemption arbitrage path between stETH and ETH — was the post-Merge withdrawal enablement at protocol level (Shapella, 2023-04), not a per-incident operator action. For OAK's purposes the case illustrates that LST/LRT depeg cascades' canonical recovery primitive is protocol-level redemption-path activation, not operator-make-whole.
OAK Techniques observed: OAK-T14.003.001(i) (LST/LRT Depeg-Cascade — Chain-Level Redemption-Absence — primary; pre-Shapella, direct stETH-to-ETH redemption at the Beacon Chain layer was not enabled, forcing all sell pressure to clear via secondary-market liquidity) + OAK-T14.003 (Restaking Cascading Risk — sub-case (b) Liquid-staking-token depeg cascade, applied here in its broader formulation that includes LSTs and not only LRTs). The proximate trigger was the Terra/UST collapse (2022-05) which caused contagion-driven sell pressure on stETH via the Anchor / bETH bridge unwinding path; the propagation channel — Curve secondary-market depeg → Aave / Compound oracle-feed propagation → looped-leverage liquidation engine — is structurally identical to the Renzo April 2024 LRT cascade at /examples/2024-04-renzo-ezeth-depeg.md. OAK-T9.001 (oracle / pricing-feed dependency) is a contributing failure mode in that lending-market oracles propagated the secondary-market discount into collateral-value calculations without a redemption-rate buffer. OAK-T14.001 is not a contributor — no slashing event fired; the depeg was driven by withdrawal illiquidity (no redemption path was available pre-Merge) and Curve-pool depth contraction, not by any underlying-validator misbehaviour. OAK-T17.002 (Liquidation-Cascade Engineering — the stETH depeg propagated through Aave and Compound oracle feeds into a cascade of looped-leverage liquidations across 442 leveraged staking positions totalling ~537,123 ETH; the cascade dynamics converted a secondary-market depeg into a mass-liquidation event, structurally identical to the pattern the T17 tactic page explicitly lists under T17.002 as the "2022-06 stETH cascade" worked example). The case does not classify under T9.005 (no rounding-error / arithmetic bug), under T10 (no bridge validator-set compromise; the Wormhole bETH bridge was the propagation path but was not itself attacked), or under T11 (no key compromise).
Attribution: pseudonymous — no external attacker; the depeg was a contagion-driven market-structure event triggered by the upstream Terra/UST collapse. The realised losses accrued to looped-leverage borrowers and to institutional holders (Celsius, 3AC) whose risk management was insufficient for the cascade. No DOJ / regulator action in connection with the stETH depeg itself; the subsequent Celsius bankruptcy and 3AC liquidation produced extensive litigation and regulator action, but not at the stETH-cascade attribution layer. The case is included because the loss was real and substantial across a documented institutional cohort, the cascade structure is the canonical pre-2024 T14.003 (b) shape, and OAK's framework treats mechanism-design-driven cohort harm as a first-class T14 surface.
Key teaching point: a liquid-staking-token without a native-redemption path (because the underlying chain has not yet enabled withdrawals) is a forced-exit-via-secondary-market design — and looped-leverage on top of it is an LST-depeg-cascade waiting to be triggered, with any upstream contagion shock as the proximate cause. The May–June 2022 Lido stETH cascade is the empirical anchor that the cascade-amplification surface predates the LRT era and is structural to the LST primitive itself when redemption is unavailable. The mitigation primitives — (1) protocol-level redemption-path activation (Shapella 2023-04 for stETH; native-redemption shipping for LRTs through 2024–2025); (2) lending-market oracle methodology incorporating redemption-rate / withdrawal-queue depth rather than pure DEX spot; (3) lending-market LTV ceilings calibrated to a configurable depeg-tail-risk buffer; (4) borrower-side awareness that LST/LRT-loop liquidation thresholds track secondary-market depth, not implied NAV — are the same primitives required for T14.003 (b) more broadly. The structural OAK lesson is that the LST/LRT depeg-cascade surface is multi-cycle: it fired in 2022 (stETH), in 2024 (ezETH), and in 2025 (stETH again, July 2025 Justin-Sun-driven cascade — see /examples/2025-07-lido-steth-aave-cascade.md); contributors should expect it to recur on any future combination of withdrawal-illiquidity + leverage-amplification + sentiment shock.

Summary

Lido is the canonical Ethereum liquid-staking protocol, deployed in late 2020 / early 2021 as a non-custodial staking pool whose participants receive stETH — an LST representing the underlying ETH staked through Lido's curated validator set. By May 2022 stETH had accumulated a multi-billion-dollar TVL with deep integration into DeFi: the Curve stETH/ETH pool was the primary secondary-market venue for stETH price discovery, with TVL above $4B at peak; Aave and Compound had integrated stETH (and wstETH, the wrapped non-rebasing form) as collateral with substantial loan books against it; looped-leverage strategies were prevalent, in which holders deposited stETH on Aave, borrowed ETH, swapped for additional stETH, and re-deposited in repeated cycles to amplify staking yield. As of the 2022-05 Terra collapse, withdrawals from Lido were not yet available — the Ethereum Beacon Chain itself did not support validator-withdrawal until the Shapella upgrade in 2023-04. The only exit path for stETH was secondary-market sale, primarily through the Curve stETH/ETH pool.

The proximate trigger for the 2022 cascade was upstream: the Terra/UST collapse on 2022-05-07 to 2022-05-16. A popular DeFi pattern at the time involved bridging stETH to Terra via Wormhole as bETH, depositing bETH into Anchor Protocol as collateral, and borrowing UST to farm Anchor's then-record ~20% deposit yield. As confidence in Terra collapsed, Anchor users rushed to unwind these positions: bETH was withdrawn from Anchor, bridged back to Ethereum, and converted back to stETH — adding sell pressure to the Curve pool. Concurrently, Three Arrows Capital (3AC) withdrew approximately $400M (~128k stETH + ~73k ETH) from the Curve pool in a single transaction on 2022-05-12, citing risk-management considerations; Celsius withdrew comparable amounts in the same window. Per Nansen's forensic analysis [nansenstethdepeg2022], the Curve pool's TVL contracted by more than 50% in three days — from ~$4.08B on 2022-05-09 to ~$1.91B on 2022-05-12 — leaving the pool's depth-at-price-tolerance materially reduced for the remainder of the cascade.

With Curve pool depth halved, secondary-market sell pressure on stETH from continuing leveraged-position deleveraging produced amplified depeg amplitude. stETH/ETH fell from ~1.00 to a low of ~0.931 on 2022-05-18 and remained at persistent discount through June [nansenstethdepeg2022]. The propagation into liquidations on Aave and Compound was immediate. Per Heimbach et al. (2023) [heimbachleveragestaking2023], 442 leverage staking positions spanning ~~537,123 ETH (~~$877M) of total exposure had been built across Aave / Compound during this period; the cohort entered a deleveraging cycle that produced ~136,069 ETH of repayments and substantial realised liquidations on the riskiest positions.

The institutional cohort was hit hardest. Celsius held one of the largest individual stETH collateral positions on Aave and approached liquidation on multiple occasions through May and June; per FinTech Collective's case-study reconstruction [fintechcollectivesteth2022], Celsius narrowly avoided full liquidation by paying down its Aave loan and withdrawing the stETH collateral entirely. Celsius subsequently filed for Chapter 11 bankruptcy on 2022-07-13, with the stETH-related liquidity stress a contributing factor in the timeline. Three Arrows Capital had built leveraged stETH-related exposure as part of its broader DeFi yield-farming book; the stETH depeg was not its primary insolvency driver (which was its Luna and GBTC exposure), but per Nansen's analysis 3AC's "lack of sound risk management coupled with excessive leverage was simply a ticking time bomb that was set off by the stETH 'de-peg.'" [nansenstethdepeg2022] 3AC entered liquidation in late June 2022.

The stETH price recovered toward NAV gradually through 2022-Q3 and Q4 as confidence returned and as the Ethereum Merge approached. Full peg restoration with arbitrage parity was achieved only with the Shapella upgrade in April 2023, which enabled validator withdrawals at the Beacon Chain level and unlocked Lido's redemption path — finally giving stETH holders a 1:1 path to underlying ETH that did not require secondary-market sale. From 2023-04 onward stETH has traded within tight peg bounds; the pre-Shapella depeg-cascade pattern is structurally retired for stETH specifically.

For OAK's purposes the case is the canonical pre-2024 T14.003 (b) anchor and a load-bearing v0.1 datapoint demonstrating that the LST/LRT depeg-cascade surface is multi-cycle (2022, 2024, 2025) and structural to any LST/LRT primitive with looped-leverage downstream and inadequate native-redemption paths. The novel OAK contribution is documenting the case as the first T14.003 (b) cohort cascade on record, distinct from the Renzo April 2024 case at /examples/2024-04-renzo-ezeth-depeg.md (which fires on the same pattern in the LRT era) and from the July 2025 stETH/Aave cascade at /examples/2025-07-lido-steth-aave-cascade.md (which demonstrates the pattern's continued operability post-Shapella when withdrawal-queue depth becomes the constrained primitive).

Timeline (UTC)

When	Event	OAK ref
2020-12 to 2021-Q1	Lido launches as Ethereum LST issuer; stETH minted against ETH deposits; Beacon Chain withdrawals not enabled at protocol level	(T14.003 (b) surface accumulates)
2021-Q2 to 2022-Q1	Curve stETH/ETH pool grows to $4B+ TVL; Aave / Compound integrate stETH as collateral; looped-leverage strategies become dominant pattern; Wormhole bETH bridge to Terra Anchor adds DeFi-yield-farming integration	(looped-leverage cohort builds)
2022-05-07 to 2022-05-16	Terra/UST collapse: upstream contagion shock; Anchor depositors rush to unwind bETH positions, bridging back to stETH and selling on Curve	(proximate trigger / contagion path)
2022-05-09 to 2022-05-12	Curve pool TVL halves: 3AC withdraws ~$400M (~128k stETH + ~73k ETH) in a single transaction on 2022-05-12; Celsius withdraws comparable amounts; pool TVL falls from ~$4.08B to ~$1.91B `[nansenstethdepeg2022]`	(T14.003 (b) liquidity-depth contraction)
2022-05-12 to 2022-05-18	stETH depeg cascade: stETH/ETH falls from ~1.00 to a low of ~0.931 on 2022-05-18; Aave / Compound looped-leverage positions enter deleveraging cycle	T14.003 (b) cascade fires
2022-05 to 2022-06	Cumulative liquidation / deleveraging cohort: 442 leverage-staking positions covering ~~537,123 ETH (~~$877M) absorb ~136,069 ETH of repayments and substantial realised liquidations on riskiest positions per Heimbach et al. `[heimbachleveragestaking2023]`	T14.003 (b) downstream extraction
2022-06	Celsius approaches Aave liquidation on multiple occasions; pays down loan and withdraws stETH collateral entirely `[fintechcollectivesteth2022]`; 3AC stETH-related leverage contributes to insolvency cascade `[nansenstethdepeg2022]`	(institutional-cohort harm)
2022-06-29	Nansen publishes "On-Chain Forensics: Demystifying stETH's De-peg" `[nansenstethdepeg2022]`; canonical forensic reconstruction of the cascade	(forensic-anchor publication)
2022-07-01 to 2022-07-13	3AC files for liquidation in BVI; Celsius files Chapter 11 bankruptcy on 2022-07-13	(institutional-cohort downstream insolvencies)
2022-Q3 to 2023-Q1	stETH/ETH peg gradually recovers toward NAV as Ethereum Merge approaches; secondary-market depth rebuilds; looped-leverage strategies re-emerge at lower aggregate scale	(pre-Shapella partial recovery)
2023-04	Ethereum Shapella upgrade: Beacon Chain validator withdrawals activated; Lido withdrawal path enabled; stETH 1:1 redemption arbitrage path becomes operational; pre-Shapella depeg-cascade surface structurally retired for stETH	(T14.003 (b) protocol-level mitigation)
2023-04 onward	Heimbach et al. publish academic analysis of leverage-staking with LSDs `[heimbachleveragestaking2023]`; Steakhouse Financial publishes LRT methodology referencing the 2022 cascade as design baseline `[steakhouselrt2024]`	(T14.003 (b) academic / methodology codification)
Continuing	T14.003 (b) cascade pattern remains operational on any LST/LRT with constrained redemption + looped-leverage downstream — the May–June 2022 Lido case is the canonical pre-2024 empirical anchor	(T14.003 (b) anchored)

What defenders observed

The cascade was triggered by an upstream contagion shock — Terra/UST collapse — but propagated through standard LST + leveraged-lending plumbing. ETH itself moved meaningfully through May 2022 but the amplitude of stETH's depeg relative to ETH was a function of Curve-pool depth contraction and forced-deleveraging-driven sell pressure, not of Lido / Beacon Chain underlying-stake stress. The defender lesson is identical to the Renzo April 2024 lesson: the trigger for an LST/LRT cascade is not load-bearing — the propagation channel (secondary-market depth + lending-market oracle + leverage liquidation engine) is the load-bearing surface.
Native-redemption-path absence is the structural prerequisite for sub-case (b) cascade — and it can be a chain-level constraint, not only a protocol-level choice. Pre-Shapella Ethereum did not support validator withdrawals at all, so Lido could not offer native redemption regardless of operator design choices. The defender lesson is that any LST/LRT issued on a chain whose underlying-stake withdrawals are not yet enabled (or are queue-bottlenecked) is a sub-case (b) surface independent of operator choice. The Renzo April 2024 case was an operator-design parallel — withdrawals could have been enabled, but were not.
Curve pool depth is the load-bearing secondary-market parameter for stETH-class LSTs and LRTs. Per Nansen's forensic analysis, the cascade fired only after the Curve pool TVL halved in three days driven by 3AC and Celsius withdrawals [nansenstethdepeg2022]. The defender lesson is that LST/LRT secondary-market depth concentration in a small number of large LP positions is itself a T14.003 (b) detection signal: a Curve / Uniswap / Balancer pool whose top-N LPs hold majority depth is one large-LP-withdrawal away from a depeg-amplitude shock.
Lending-market oracle methodology that anchors to DEX spot is the load-bearing T9.001 contribution to T14.003 (b), and the 2022 cascade is its earliest empirical anchor. Aave and Compound's stETH oracle methodology in 2022 anchored to DEX-spot (with various smoothing primitives); the secondary-market discount propagated directly into liquidation-trigger calculations. A redemption-rate-aware oracle would have produced a much smaller depeg signal — but pre-Shapella the redemption rate itself was not directly observable on-chain. The defender lesson is that LST/LRT oracle methodology is a T14.003 (b) mitigation surface, but its design depends on the redemption-path's on-chain observability.
The recovery posture (no operator-make-whole, gradual price recovery as upstream shock dissipates and as protocol-level redemption is enabled) is the realistic 2022 ceiling. Lido did not compensate the looped-leverage cohort; Aave / Compound did not adjust their oracle methodologies retroactively; the price recovered through arbitrage and (eventually) protocol-level redemption activation at Shapella. The defender lesson is that LST/LRT depeg-cascade incidents have a structural ceiling on per-incident recovery — the canonical recovery primitive is protocol-level redemption-path activation, not operator-make-whole. Contributors writing future T14.003 (b) entries should not over-claim recovery.

What this example tells contributors writing future Technique pages

T14.003 sub-case (b) is multi-cycle and pre-dates the LRT era. The May–June 2022 Lido stETH cascade is the earliest documented empirical anchor for the pattern that would recur with Renzo ezETH in April 2024 and with stETH again in July 2025. Contributors writing future T14.003 (b) entries should treat the cascade pattern as steady-state for any LST/LRT primitive with constrained redemption + looped-leverage downstream, and reference all three cases as the canonical worked-example trio.
Chain-level redemption-path constraints are a separate sub-class within T14.003 (b) from operator-design redemption-path constraints. The 2022 stETH cascade fired because Beacon Chain withdrawals were not yet enabled (chain-level constraint, no operator could ship redemption); the 2024 ezETH cascade fired because Renzo had not yet shipped native withdrawals (operator-design choice). Both produce the same cascade structure; the mitigation differs (chain-level requires protocol upgrade; operator-level requires operator decision). Contributors writing future T14.003 (b) entries should preserve the distinction.
The institutional-cohort harm (Celsius near-liquidation, 3AC contributing-factor) is a distinct T14.003 (b) feature absent from the 2024 retail-borrower cohort cases. Pre-2022, LST exposure was concentrated in CeFi lenders and crypto-native hedge funds; the 2022 cascade's downstream cohort included institutional-scale balance sheets. The 2024 ezETH cohort by contrast was retail-dominant looped-leverage. Contributors writing future T14.003 (b) entries should report the cohort composition explicitly — it determines whether the downstream impact propagates into broader CeFi insolvency (2022 pattern) or remains contained at the retail-borrower lending-protocol layer (2024 pattern).
Curve pool depth concentration is itself a T14.003 (b) detection signal independent of any LST issuer. The Nansen forensic finding that 3AC's single $400M withdrawal halved the Curve pool's TVL is the empirical anchor that secondary-market depth concentration is a structural T14.003 (b) precondition. Contributors writing future T14.003 (b) entries should report top-N LP concentration on the dominant secondary-market venue as a first-class detection signal.
Pre-Merge stETH and post-Merge LRTs share the cascade structure but differ in their structural fix. The structural fix for pre-Merge stETH was the Ethereum Merge + Shapella (chain-level protocol upgrade); the structural fix for LRTs is per-issuer native-redemption-path activation (operator-level choice). Contributors writing future T14 mitigation entries should preserve the distinction explicitly: chain-level fixes retire the cascade for the LST cohort on that chain; operator-level fixes retire the cascade only for that LRT issuer.

Public references

[nansenstethdepeg2022] — Nansen, "On-Chain Forensics: Demystifying stETH's De-peg" (2022-06-29); canonical forensic reconstruction of the cascade with primary numerical figures (Curve pool TVL contraction, 3AC + Celsius withdrawals, depeg amplitude).
[fintechcollectivesteth2022] — FinTech Collective, "stETH Depegging: A Case Study of Cascading Events" (2022); structural-analysis source for Celsius near-liquidation and the cascade-event framing.
[heimbachleveragestaking2023] — Heimbach, Schertenleib, Wattenhofer, "Leverage Staking with Liquid Staking Derivatives (LSDs)" (IACR ePrint 2023/1842); academic analysis with primary numerical figures on leverage-staking cohort size.
[coindesksstethterra2022] — CoinDesk, "Nansen Report Shows Links Between Terra Collapse and stETH 'De-peg'" (2022-06); secondary coverage source.
[steakhouselrt2024] — Steakhouse Financial steakLRT methodology and LRT risk-disclosure framework; references the 2022 stETH cascade as design baseline.
[gauntletrestaking2024] — Gauntlet 2024 restaking-economy analysis; LRT collateral-health methodology framework with 2022 stETH cascade as reference event.
[alexanderleveragedrestaking2024] — Carol Alexander et al., "Leveraged Restaking of Leveraged Staking: What are the Risks?" (SSRN 2024); pre-event formalisation of the depeg-and-liquidation cascade amplification, references 2022 stETH cascade.

Citations

[nansenstethdepeg2022] — Nansen forensic write-up; primary on-chain numerical anchor.
[fintechcollectivesteth2022] — FinTech Collective case study; structural-analysis source.
[heimbachleveragestaking2023] — Academic analysis; primary cohort-size numerical anchor.
[coindesksstethterra2022] — CoinDesk; secondary coverage.
[steakhouselrt2024] — Steakhouse Financial LRT methodology.
[gauntletrestaking2024] — Gauntlet restaking-economy analysis.
[alexanderleveragedrestaking2024] — Alexander et al. leveraged-restaking risk paper.

Discussion

The May–June 2022 Lido stETH depeg-and-liquidation cascade is OAK's pre-2024 anchor for T14.003 sub-case (b) and the canonical multi-cycle datapoint for the LST/LRT depeg cascade surface. The case sits as the earliest in a documented cascade-pattern trio: 2022 Lido stETH (this case), 2024-04 Renzo ezETH (/examples/2024-04-renzo-ezeth-depeg.md), and 2025-07 stETH / Aave / Justin-Sun-driven cascade (/examples/2025-07-lido-steth-aave-cascade.md). All three share the structural form — secondary-market sell pressure → DEX pool depth contraction → lending-market oracle propagation → looped-leverage liquidation cascade — but each illustrates a different constrained primitive: pre-Merge withdrawal unavailability (2022); operator-design withdrawal blocking (2024); withdrawal-queue depth saturation under exit-queue congestion (2025).

The pre-event design-document framing in 2022 was thin relative to 2024–2025: Steakhouse Financial's LRT methodology [steakhouselrt2024] and Gauntlet's restaking-economy analysis [gauntletrestaking2024] were not yet published; Heimbach et al.'s academic analysis [heimbachleveragestaking2023] was published in response to the 2022 cascade. The case is therefore the generative empirical anchor for the cascade pattern — the event that made the design-document framing canonical for the 2024 LRT era.

The institutional-cohort harm dimension distinguishes the 2022 case from the later cases. Celsius's near-liquidation on Aave was at institutional scale; 3AC's stETH-related leverage was a contributing factor (not primary cause) in its insolvency. The 2024 ezETH cohort by contrast was retail-dominant looped-leverage with no institutional-scale insolvency downstream. Contributors writing future cross-Tactic mappings (T14.003 → T11 / T8 institutional-failure surfaces) should treat the 2022 stETH case as the empirical anchor that LST/LRT cascade can propagate into broader-CeFi-insolvency surface when institutional balance sheets carry the looped-leverage exposure.

The recovery posture — gradual market-clearing through 2022-Q3/Q4, with structural fix delivered only at Shapella in April 2023 — is informative for OAK's T14.003 (b) Mitigations layer: the canonical recovery primitive for a chain-level-redemption-constrained LST cascade is protocol-level redemption-path activation, with no operator-side acceleration possible. Contributors writing future T14.003 (b) entries on chains with constrained withdrawal paths should treat protocol-level redemption activation as the structural cure and per-issuer mitigation as containment.

For OAK's broader credibility, including the 2022 Lido stETH case as a T14.003 (b) anchor closes a temporal gap: the v0.1 corpus's T14.003 cases were 2024–2026 only prior to this addition. The 2022 case operationalises T14.003 (b) as a multi-cycle pattern with a documented pre-2024 empirical anchor and provides contributors writing future T14.003 (b) entries with the canonical generative reference.