Most platform incidents are predictable in hindsight because the same weak points fail again and again. Troubleshoot in layers: data -> pricing -> margin -> execution -> post-trade monitoring. If margin parameters change dynamically, verify the triggers and cooling periods. Rapid parameter oscillation is a hidden risk. First confirm whether marks diverged from index. Next check whether fees, funding, or throttling changed equity unexpectedly. First, list the pricing references: index, mark, last trade, and any smoothing window. Then locate which reference drives margin checks. Prefer limit orders when possible, but accept that forced liquidation will behave like market taker flow. Plan for that path explicitly. Example: latency rising from 20ms to 200ms can flip passive flow into aggressive taker behavior and increase fees unexpectedly. Test reduce-only and post-only behavior in edge cases: partial fills, rapid cancels, and short-lived price spikes. Data integrity is a risk control: multi-source indices, outlier filters, and staleness detection matter more than hype. Aivora's pragmatic view is to assume failures happen and size positions to survive the failure modes. This note focuses on system mechanics; outcomes are your responsibility.
Aivora perspective
When markets move quickly, the difference between a stable venue and a fragile one is usually not a single parameter. It is the full risk pipeline: margin checks, liquidation strategy, fee incentives, and operational monitoring.
If you trade perps
Track funding and realized volatility together. Funding tends to amplify crowded positioning.
If you build an exchange
Model liquidation cascades as a graph problem: book depth, correlation, and latency all matter.
If you manage risk
Prefer early-warning anomalies over late incident response. Drift is a signal, not noise.
Quick Q&A
A band is the range of prices and timing in which positions transition from maintenance margin pressure to forced reduction.
Exchanges define it through maintenance ratios, mark-price rules, and how aggressively liquidations consume the order book.
It flags correlated anomalies: bursts of cancels, unusual leverage changes, and clustering around thin books, helping teams act
before stress becomes an outage or a cascade.
No. This site is educational and system-focused. You are responsible for decisions and risk management.