USDe dislocations were a late symptom, not the spark; depth thinned first, and the “deep pool” undercut earlier and harder than smaller venues.
Risk channel congestion: ReduceOnly/Close orders repeatedly hit 503 / -1008 / -4118 / -2022, so risk couldn’t shrink on schedule.
PM design gap: “Close All” exists in the UI; there’s no public one-shot API to flatten cross-products to a chosen asset—system path fell back to human path at peak stress.
Priority inversion: matching/risk capacity was effectively tilted toward liquidation flow, not self-rescue flow.
Local methodology (index/bands) amplified prints and leaked into mark/liq thresholds.
Insurance vs. ADL sequencing didn’t provide a strong buffer; ADL stepped across cohorts (from >2x toward ~1x).
Time profile: same-second liquidations, queued stops, delayed warnings for roughly a hundred-plus minutes—classic congestion.
Outcome: deeper slides in the deepest pool, index leakage, cross-venue follow-through—a death spiral.
Aftermath (compensation, wrapped-asset pricing/index changes) tacitly admits fragility in prior parameters/methods.
Call: The primary driver wasn’t “greed/fear,” but safety rails failing when needed most. Market risk is ours; infrastructure risk shouldn’t be.
Our position
Truth as we see it: Oct-11 reads as an infrastructure incident—failed last-resort risk controls + venue-local methodology amplified + price-source leakage → system-wide ADL.
Tone: Facts over emotion; no doxx. If someone has the full per-second platform logs, prove us wrong. Until then, this is the most reasonable reconstruction.