Summary
On October 10, 2025, the cryptocurrency market experienced the largest liquidation cascade in history. Over $19.2 billion in leveraged positions were liquidated in a matter of hours. This was approximately 9x larger than the FTX collapse and 18x larger than the COVID crash of March 2020.
During this unprecedented volatility, FiRM processed 79 liquidations across three markets: CVX (72), CRV (5), and cvxCRV (2), clearing approximately $717,000 DOLA - one of the largest 24-hour liquidation periods in the protocol's history. The protocol maintained full operational stability throughout the event, DOLA retained its peg, and FiRM TVL and total outstanding debt was largely affected due to the stable nature of the collateral deposited in the protocol.
Volatile assets listed as collateral on FiRM experienced a rapid price collapse, some exceeding 70%. CVX dropped 71% in just 27 minutes, from $2.94 to $0.85, and was particularly relevant given its market’s size prior to this event. The extreme velocity of the CVX price collapse resulted in approximately $110,000 DOLA in bad debt, isolated entirely to the CVX market. This bad debt has been repaid as of October 16, 2025, through the generous contribution of CVX founder C2tp, core contributor Winnie, and the Convex Treasury, who jointly covered the debt. FiRM remains free of bad debt and fully operational, and we extend our gratitude to Convex for their partnership and support.
Technical Analysis: Oracle Behavior Under Stress
The RWG conducted a detailed analysis of the Chainlink CVX/USD oracle performance during the liquidation window (20:00 - 23:00 UTC, October 10). The findings reveal that while the oracle operated correctly by design, network congestion during the historic liquidation event created latency that amplified the impact on liquidations.
Key Oracle Metrics (Oct 10, 20:00 → 22:10 UTC - Crash Window)
Total Rounds Observed: ~48 updates
Average Update Interval: ~1.54 minutes
Maximum Update Interval: 10 minutes 36 seconds
Lagged Updates (>120s and Δ>2%): 10 events (~21% of rounds)
Total Duration of Lag Events: ~40.6 minutes
Largest Price-Change Cluster: 21:17-21:25 UTC (~58% decline)
Oracle Deviation Threshold Behavior
The CVX/USD Chainlink feed is configured with a 2% deviation threshold, meaning the feed should push a new round once the aggregated off-chain price differs by ≥2% from the last on-chain value. During the October 10 volatility window, the feed did not update consistently at the 2% threshold:
Several price deviations of 3-5% persisted for 6-10 minutes before the next update
Updates frequently arrived in clusters, each capturing multiple unreported price moves
On-chain publication lagged behind off-chain detection, likely due to network congestion and node-level throttling across CVX/USD and CVX/ETH feeds during the cascade.
This behavior created step-wise repricing instead of smooth tracking. The 2% deviation rule triggered off-chain as designed, but on-chain publication was delayed, compressing multiple minutes of market movement into single on-chain adjustments.
Correlation with FiRM Liquidations
The most critical period occurred during the crash epicenter between 21:17-21:25 UTC (8 minutes), when CVX/USD fell from approximately $2.30 to $0.98—a 58% decline compressed into five consecutive oracle rounds. Each of these updates exceeded 5-25% deviation, with the largest single-round move at -25.1%. Update intervals during this window were irregular (36-204 seconds), with three lagged updates exceeding 2 minutes.
This 8-minute crash epicenter aligned precisely with FiRM's primary liquidation cluster on October 10. The delayed on-chain repricing intensified collateral markdown density, causing multiple positions to breach liquidation thresholds simultaneously and creating the bulk of the bad debt accounts in a concentrated timeframe.
A separate analysis carried out on the CVX/ETH feed showed nearly identical compression during the 21:23-21:27 UTC window, where the feed dropped approximately 30% across just 3 blocks (36 seconds). At this velocity - over 2% per minute sustained across 27 minutes - even optimally configured parameters face challenges as liquidators must process transactions faster than the market is repricing collateral.
After approximately 22:10 UTC, both price volatility and update intervals normalized, confirming the issue was temporary congestion rather than systemic oracle failure.
Oracle Integrity Assessment
The Chainlink CVX/USD oracle maintained full integrity throughout the event:
All rounds settled sequentially on-chain with no data gaps or invalid rounds
Feed performance degradation was latency-driven, not failure-driven
Responsiveness normalized as volatility subsided
The oracle updated correctly by rule, but not promptly by market pace
This demonstrates oracle latency under unprecedented stress rather than misconfiguration or systemic failure.
Context: RWG Risk Monitoring Framework
RWG maintains weekly and monthly risk observer checklists that track liquidity metrics, trading volumes, and market depth across all collateral assets listed on FiRM. This systematic monitoring framework has proven effective in identifying risk early - for example, in July 2025, ongoing monitoring identified concerning liquidity decline in the cvxCRV and st-yCRV markets. RWG brought forward a proposal to the DAO, which approved increasing liquidation factors to 100% for both markets. These markets experienced zero bad debt during the October 10 event. Additionally, the RWG has worked with Chainlink in the past to lower deviation thresholds on stablecoin price feeds - successfully reducing deUSD and USR feeds down from to 2% and 1% respectively to 0.5% deviation thresholds, providing tighter price tracking for stablecoin collateral during volatile periods.
In the days, weeks, and months leading up to the October 10 event, CVX exhibited healthy liquidity conditions. A snapshot taken immediately before the crash showed the cvxETH pool on Curve maintaining ~$12.5M TVL, with 7-day low trading volume nearly doubling from $8.2M to $14.4M. There were no warning signs of deteriorating liquidity or structural weakness that would have triggered parameter adjustments through standard monitoring processes.
Understanding FiRM's Liquidation Factor Modeling
The RWG uses liquidation factor modeling to balance two critical objectives: setting parameters conservative enough to handle extreme liquidation scenarios while remaining competitive with other lending protocols to attract users. This modeling accounts for gas costs, price volatility, and liquidation profitability to ensure liquidators can successfully clear underwater positions before they become bad debt.
All bad debt positions resulted from the 21:17-21:25 UTC crash epicenter, where the 58% price decline in 8 minutes outpaced on-chain price updates and liquidation execution at the 60% liquidation factor setting.
While more frequent model updates during volatile bull markets would provide additional safety margin, the core issue in this event was the unprecedented combination of velocity and magnitude that exceeded reasonable modeling assumptions, compounded by oracle latency that compressed price updates into step-wise adjustments.
What Worked: Defense in Depth
Several risk controls performed exactly as designed during this extreme stress test:
Liquidator Participation: Ten unique liquidators participated in the 53 liquidation events on October 10, with the primary liquidator handling 35 events (~66%) across all three affected markets. This competitive participation ensured liquidations were processed successfully despite network congestion.
Markets with 100% Liquidation Factors: The cvxCRV and st-yCRV markets (both configured with 100% liquidation factors following July 2025 parameter updates) experienced zero bad debt despite exposure to similar volatility.
Stablecoin LPT Oracle Protections: Our stable LP oracle design - chainlink feeds paired with PPO and $1 hardcoded DOLA cap - functioned as intended and as a result no liquidations occurred for stablecoin LP markets despite, for example, USDe experienced secondary market volatility, dropping to $0.62 on centralized exchanges while maintaining $0.9929 on Chainlink feeds or crvUSD momentarily depegging upwards to $1.09.
Liquidation Mechanisms: All liquidation mechanisms operated as intended - the $110k bad debt represents the narrow gap between liquidation execution speed and the extreme price collapse velocity at 60% LF configuration, amplified by oracle latency.
Immediate Actions and Mitigation
Based on comprehensive analysis of this event, RWG is implementing the following changes:
1. Market Parameter Safety Update
The RWG will bring forward a proposal to the DAO to revise safety parameters across volatile markets where deemed appropriate, with a focus on CVX and CRV. Proposed changes may include revisions to either/both collateral factors and liquidation factors, supported by data and to provide stronger protection against extreme volatility events. These parameter adjustments represent immediate protective measures.
2. Oracle Infrastructure Review
RWG is initiating discussions with Chainlink to understand oracle behavior under extreme network congestion and explore potential improvements:
Evaluation of whether lower deviation thresholds could improve responsiveness during rapid price movements
Assessment of redundant oracle architectures, potentially incorporating pull-based oracle functionality for liquidators that validates against Chainlink's push-based feeds
Investigation of whether additional data providers (Redstone, Pyth, Chronicle) could provide real-time position pricing supplements during extreme events
3. Technical Analysis Framework
RWG has developed a custom oracle analytics tool in Python for post-event analysis and ongoing monitoring:
Web3.py integration with Ethereum mainnet to fetch historical Chainlink round data
Automated lag detection logic identifying rounds where update intervals exceed thresholds
CSV integration combining oracle data with FiRM liquidation records to align timing
High-resolution visualization tools for oracle responsiveness analysis
This framework enables rapid forensic analysis of future events and supports data-driven parameter adjustments.
4. Liquidation Engine Enhancements (Medium-Term)
RWG and PWG will explore upgrades to FiRM's liquidation architecture, including:
Sliding scale liquidation factors that adjust automatically based on real-time volatility
Soft liquidation mechanisms similar to those employed by other lending protocols
Hybrid oracle systems that enable liquidators to access real-time pricing while maintaining Chainlink feeds as the authoritative source
Bad Debt Resolution
The $110,310 in bad debt was distributed across 14 positions in the CVX market. This amount represents 0.1% of FiRM's total outstanding debt and was incurred during the largest liquidation cascade in cryptocurrency history.
As of October 16, 2025, this bad debt has been fully repaid through a joint contribution from CVX founder C2tp, core contributor Winnie, and the Convex Treasury. We are deeply grateful to Convex for stepping forward to cover this bad debt. Their support demonstrates the strength of the partnerships FiRM has built and their confidence in the protocol's infrastructure. Inverse Finance maintains a consistent track record of bad debt reduction - last month alone, Treasury repaid over $300,000 in legacy bad debt as part of its ongoing commitment to protocol health.
Looking Forward
This event represented an unprecedented stress test for decentralized lending infrastructure across the entire cryptocurrency ecosystem. While the $110k in bad debt is an outcome we take seriously, it represents a marginal amount compared to our books and of FiRMs current outstanding debt.
This event highlights the importance of ongoing parameter vigilance and the value of conservative configurations for volatile assets. The proposed liquidation factor increases will be brought forward through governance for DAO approval in the coming days.
RWG will continue monitoring for potential contagion effects in the coming weeks and coordinate with partner protocols to share insights from this event. FiRM remains fully operational, solvent, and stable across all markets. We remain committed to transparency, continuous improvement, and maintaining FiRM as a secure and reliable lending infrastructure for the Inverse Finance ecosystem.
