Depeg Tracker + DEWS Changelog

Impact Notes

• Pending rows are now refreshed while they wait for confirmation, preserving first-seen timestamps but updating last-seen and worst-seen state for the active direction
• Pending confirmation now treats opposite-side secondary evidence as contradiction instead of support, and native-quote recovery no longer persists contradictory `recovery_price` values
• DEWS divergence now ignores fresh-but-thin DEX rows unless they pass the same `$1M` live depeg trust gate, while the repair path refreshes current rows and prunes unrecomputable daily history from the Mar 9, 2026 trust-floor boundary onward

Impact Notes

• Historical native-fiat replay now uses daily points plus a two-point confirmation window across 36 hours before opening normal non-USD fiat backfill events
• Extreme single-point native crashes of 5,000 bps or more are still preserved even when the native historical confirmation rule would otherwise suppress them
• Broad non-USD backfill repairs can now use authenticated CoinGecko market-chart transport consistently through the admin/backfill path, reducing false rebuild drift during large historical cleanups

Impact Notes

• PegScore now uses a curated launch date as the age anchor when one is available, falling back to earliest supply-history coverage only when no launch date is curated
• Older tracked assets with late `supply_history` coverage no longer appear artificially young in the peg-score window
• BRZ now anchors peg-score tracking to its July 19, 2019 launch date instead of a March 2025 supply-history coverage artifact

Impact Notes

• Fresh direct native-peg quotes can now veto, sustain, or resolve live depeg state for supported EUR/CHF/GBP/JPY/SGD/AUD/CAD/BRL/IDR/TRY/ZAR/PHP/MXN/RUB/CNH-style fiat pegs instead of remaining effectively BRL-only
• Pending confirmation prefers that same direct native quote first whenever CoinGecko exposes the matching fiat pair, reducing false confirmations from USD/FX reference drift
• Historical backfill now replays supported non-USD fiat assets against direct native fiat price history and a native `1.0` peg when that series exists, which removes large classes of synthetic backfill events caused by USD/FX mismatch

Impact Notes

• BRZ-style depegs can now be vetoed or resolved by a fresh direct BRL quote when the USD/FX-derived signal is back inside threshold or points the other way
• Pending confirmation prefers the fresh direct native-peg quote over a weaker derived USD cross-check when CoinGecko exposes that native pair
• Prevents thin BRL reference mismatches from opening, sustaining, or confirming false depeg rows while still preserving genuine native-peg stress

Impact Notes

• Thin fiat peg groups with fewer than 3 live contributors now require cached FX fallback before they can open, update, or resolve live depeg state
• Fresh non-cached multi-source primary agreement can now close an already-open live event after recovery even if that source mix is still too soft to open new events directly
• Prevents BRL-style peer-median false positives from both opening new live rows and lingering as active depegs after the peg reference normalizes

Impact Notes

• Coins with >= 7 days of mint/burn history now keep the flow signal available even when the latest 24-hour bucket is quiet
• A zero-flow day now contributes zero flow stress instead of silently redistributing the flow weight away from the score
• DEWS cron assembly now reflects the documented mint/burn-baseline semantics for mature tracked coins

Impact Notes

• DEX-assisted live recovery now requires at least two corroborating protocol-level DEX groups inside threshold, not just one trusted aggregate row
• Large challenger pools can veto ambiguous-primary DEX recoveries when they still show the old depeg direction
• New-event suppression and other aggregate-DEX-driven state mutations now share that stricter corroboration policy, reducing synthetic split events on chronically depegged coins

Impact Notes

• Same-direction DEX disagreement is now advisory for ongoing events instead of forcing a synthetic recovery boundary
• Open depeg rows remain continuous until the standard recovery path confirms a sub-threshold resolution
• Prevents repeated event fragmentation on chronically depegged assets when aggregate DEX pricing briefly snaps back toward peg

Impact Notes

• Blacklist signal coverage is now derived from the shared supported blacklist symbol set, so PYUSD and USD1 receive the same DEWS blacklist input treatment as USDC, USDT, PAXG, and XAUT
• Thin non-USD DEWS divergence inputs now use cached `fxFallbackRates` from the stablecoins payload instead of relying only on sparse peer medians
• DEWS input semantics now match the documented blacklist coverage and the live depeg / peg-summary FX-reference path

Impact Notes

• Bootstrap grace now ends after the first successful DEWS publication via a dedicated `dews:bootstrap-complete` sentinel instead of piggybacking on stablecoins-cache freshness
• Only explicitly optional missing tables remain bootstrap-allowed before first success; core dependencies no longer inherit that grace
• Fresh `dex_liquidity` is now required for publication, with rows older than 2 hours treated as a hard degraded source failure

Impact Notes

• Opposite-direction live depeg rows no longer remain active just because the correcting primary price is still confirm_required
• Direction flips from cached, fallback, low-confidence, or stale primary inputs now close the stale live row and insert a replacement pending candidate
• Public active-depeg state stops claiming the wrong side of the peg while confirmation catches up on the corrected move

Impact Notes

• Minimum tracking threshold reduced from 30 to 7 days — coins receive a composite peg score after their first week
• Scores based on 7–30 days of data are labelled 'Early score' in the hero card (amber text with tooltip)
• Report card peg-stability dimension is now rated from day 7; NR only appears for coins with < 7 days of history

Impact Notes

• Cached, fallback, low-confidence, and stale primary prices now require confirmation before they can open or close live depeg state
• Extreme moves no longer get dropped just for crossing the old <0.5x or >2x peg guardrail; they enter a dedicated confirmation lane instead
• peg-summary now exposes price provenance and trust state, and the depeg page consumes backend freshness metadata plus real event-history pagination

Impact Notes

• Depeg suppression/confirmation now requires fresh DEX rows with >= $1M aggregate source TVL
• UI-facing dexPriceCheck exposure now requires fresh data with >= $250K aggregate source TVL
• Thin DEX rows can no longer veto new depeg events or promote pending confirmations on their own

Impact Notes

• coinTrackingStart now returns null when both firstSeen and events are absent
• computePegScoreWithWindow now yields null pegScore for insufficient-history coins
• Prevents false perfect-score outcomes on sparse or incomplete datasets

Impact Notes

• Tracking start formalized as max(firstSeen, fourYearsAgo) with earliest-event fallback
• Per-event severity now applies max(duration penalty, magnitude floor)
• Active-depeg penalty steepened to max(5, absBps/50), capped at 50

Impact Notes

• Added 8th DEWS sub-signal: yield anomaly (weight 0.05)
• Introduced systemic amplifier: DEWS boosted up to +30% when PSI < 75
• Cron now reads yield_data.warning_signals and latest PSI sample before scoring

Impact Notes

• avg_pool_stress is now consumed as native 0-100 (removed erroneous x100)
• Pool component returned to intended weighting and magnitude
• Reduced false high-band classifications caused by scale inflation

Impact Notes

• Introduced DEWS base model with 7 sub-signals and weighted-redistribution scoring
• Threat bands established: CALM, WATCH, ALERT, WARNING, DANGER
• compute-dews cron writes rolling stress_signals and daily stress_signal_history snapshots

Impact Notes

• Lookback boundary corrected from min(firstSeen, fourYearsAgo) to max(...)
• Peg-time and severity now computed against realistic coin lifetime bounds
• Young-coin scores became less artificially inflated

Impact Notes

• Pending rows with invalid peg_reference are dropped before confirmation math
• Detection now rejects non-finite peg references before bps computation
• Pending confirmation mutations are batched atomically for consistent state transitions

Impact Notes

• >= $1B coins now route through depeg_pending before promotion to depeg_events
• Secondary agreement uses CoinGecko and/or DEX with a 50% threshold bar
• sync-stablecoins now runs detectDepegEvents and confirmPendingDepegs sequentially each cycle

Impact Notes

• computePegScoreWithWindow introduced a 4-year tracking cap
• Detail-page peg scores became explicitly lookback-bounded
• Boundary logic was later corrected in v3.2 for firstSeen direction

Impact Notes

• Severity penalty changed from sqrt-based to linear (peakBps/100 scaling)
• Added spreadPenalty from event-magnitude standard deviation (cap 15)
• Composite became: 0.5*pegPct + 0.5*severity - activePenalty - spreadPenalty

Impact Notes

• Non-USD depeg threshold raised to 150 bps (USD remained 100 bps)
• Cleanup migration removed legacy non-USD events below 150 bps
• Ongoing events auto-close after sustained DEX disagreement (30m+, >=$1M TVL)

Impact Notes

• Added active-depeg penalty to peg score, then introduced a minimum floor
• Merged overlapping depeg intervals to prevent pegPct double-counting
• Detection now closes orphan open events and tracks open state by event ID

Impact Notes

• Launched computePegScore baseline (peg time + severity blend)
• Introduced detectDepegEvents cron pipeline with live open/close/update logic
• Added duplicate-open-event merge and new-event DEX disagreement suppression