docs(lore-0172): USDT/USDC candles close at ~0.14, not ~1.00

[karczuRF]

Found while verifying the 0165 deploy on prod. USDT was 0165's control asset — the peg arm must not flatten it — and that check passed. But the values it returned are impossible.

Measured

price_usd_series, USDT @ canonical issuer:  method=traded  rows=891  avg_close=0.1923
price_ohlcv_1d, asset_id 111, quote=USDC:   891 candles, avg 0.1923, min 0.0912, max 0.4915
by source:                                  sdex, 132,685 trades, 491,984 USDT volume

Two dollar-pegged assets cannot trade at 0.09–0.49 against each other.

close_usd ≈ close_native is correct — a USDC-quoted candle takes the stablecoin-direct tier, so close_usd = close × $1. The defect is in close itself.

Not caused by 0165

Arm B contributes 0/0 wherever sum(w) > 0, so every one of these values is arithmetically identical to what the old view published. Pre-existing; nobody had looked at USDT.

Two hypotheses tested and falsified — recorded so they aren't re-run

1. 0139 asset_id collision ❌ — asset_id = 111 resolves to exactly one identity.
2. Misattributed XLM prices ❌ — 0.142572 vs 0.163999 on the same day; close in magnitude, never equal.

Also a query gotcha worth keeping: price_ohlcv_1d holds one row per source, so a naive join to XLM/USDC fans out ~4× per day.

Measured in passing, and worth carrying to 0139

3,281 asset_ids serve 6,568 natural identities. 0139 currently records the collision as a schema-level suspicion with no blast radius; this is the first measurement I'm aware of.

Why it matters

close_usd is what BE multiplies into TVL. If USDT positions value at ~$0.14 instead of ~$1.00, every USDT-legged pool is understated ~7×. BE's CSV has 102 priceable USDT pools — pools that look healthy and are silently wrong, which is worse than the never-priced ones 0165 just fixed.

⚠️ It also means USDT is not a trustworthy control for peg work until resolved — 0165 used it as one.

The task carries four concrete next probes (writer-side id assignment, raw-trade vs candle agreement, orientation/scaling, other USDT pairs).