v1.0.0: seven-path backtest, calibrated liquidator and borrower behavior, per-asset haircuts

First tagged release of Spine, matching the 2026-09-19 writeup (Phase G).

What is in it

• Live lane: hourly fetch of every Morpho Blue position and liquidation on Coinbase's nine Base markets, Coinbase Smart Wallet tagging via Multicall3, CEX L2 depth and DEX quoter capacity, summarized to a static dashboard on GitHub Pages.
• Model lane: Chainlink oracle paths for the three lived stress events (Oct 2025, Feb 2026, Jun 2026); historical books rebuilt from transaction events; liquidator latency, bonus, and throughput calibrated on 8,000+ cbBTC liquidations; a two-parameter borrower-response fit that reproduces all three events within 15%.
• Backtest: today's book replayed through seven crash paths (Mar 2020, May 2021, FTX, Aug 2024, Oct 2025, Feb 2026, Jun 2026) in a numpy liquidation-queue simulation with finite DEX and CEX capacity, three capacity scenarios, LLTV and capital sweeps, held-at-the-low sensitivity, alt-market replays, and a book-size sweep to 32x.

Headline: on a March 2020 path at today's 86% LLTV, realized loss is $9.2M (0.6% of supply) but $158M (10%) sits underwater and unserved at the trough. The LLTV buys about a point of supply per step down; liquidator capital buys three to six. Recommendation: a committed backstop liquidator first, then 80% LLTV for cbBTC, 77% for WETH, per-asset caps for the alts. Full argument in WRITEUP.md.

Every table in the writeup is printed by a script in this repo from data/.