A third path: Tiered LTV by derivative depth + productivity constraints
Both Scenario 1 and Scenario 2 address loss allocation after the fact. Neither addresses why this loss was structurally inevitable — and will happen again unless the underlying parameter framework changes.
The root failure: In January 2026, Proposal 434 raised rsETH’s LTV to 93% in E-Mode. A 7% safety buffer on a depth-3 derivative asset (ETH → stETH → EigenLayer → rsETH → bridge) could not absorb ANY of its failure modes: staking slashing (2-5% depeg), AVS failure (10-30%), or bridge exploit (as we saw: 73.54% shortfall). The buffer was calibrated for normal market volatility, not for structural failure of the derivative chain.
On Scenario 1 vs 2: Scenario 2 is structurally closer to correct. The bridge adapter contained 40,373 rsETH backing 152,577 claims — mainnet rsETH IS fully backed. Socializing bridge-specific failure to mainnet holders creates moral hazard: no price signal discouraging risky bridge configurations. However, L2 WETH depositors who never interacted with rsETH became collateral risk through Aave’s listing decision — they deserve a DAO treasury backstop.
My proposal — tiered LTV by derivative depth:
| Tier | Asset type | Max LTV (E-Mode) | Rationale |
|---|---|---|---|
| 1 | Native (ETH, USDC) | 80-85% | Direct redemption, no intermediary failure |
| 2 | 1st-order derivative (stETH, rETH) | 75-80% | Single rehypothecation, withdrawal queue risk |
| 3 | 2nd-order derivative (rsETH, eETH) | 65-70% | Restaking slashing + base staking risk |
| 4 | Bridged rehypothecated (rsETH on L2) | 50-55% or ineligible | Bridge exploit = instant total loss, not gradual depeg |
Additionally — productivity constraints for depth 3+:
Bridged restaking tokens should not be depositable as collateral at all. They should be holdable, transferable, redeemable — but not leverageable. Each additional leverage cycle on a depth-3 asset amplifies contagion without adding economic value. The math: depth × LTV^n produces the leverage multiplication that created the 45:1 contagion ratio we observed ($292M exploit → $13B TVL loss).
On the Umbrella Module: Do not pause it. The explicit promise to stkAAVE holders was automatic slashing under defined conditions. A discretionary governance override sets a precedent that every future parameter is subject to retroactive change — destroying the credibility that makes the Safety Module attractive to stakers in the first place.
On accountability sequencing: KelpDAO + LayerZero bear primary liability (1-of-1 DVN is negligent infrastructure). Remaining L2 losses should be backstopped by DAO treasury for WETH depositors who never chose rsETH exposure. The Umbrella Module activates per its design for any remaining shortfall.
I’ve published a full analysis with framework evidence supporting this position: The rsETH Incident Proves: Rehypothecated Assets Need Higher Collateral Ratios
The underlying methodology — including the Collateral Derivative Concentration formula, Redemption Mismatch computation (7,000x ratio for bridged LSTs), and Bridge Infrastructure scoring — is open source at docs/frameworks/24-systemic-dependency-mapping.md.
Arbitrum Seizes ~37% Ilicit ETH