One Year of Executed Order Flow Auction Trades Across CoW Protocol, UniswapX, and 1inch Fusion

v2.0.5 replaces replication_code.tar.xz only; no data file changes and no paper number changes. It completes the audit of numeric literals sitting in hardcoded print() text rather than in interpolated expressions. All three such figures that either paper quoted are now computed: c09 printed an auction-venue span of "194 to 449" matching no computed quantity and now computes 184.3 to 452.4; c10 printed the UniswapX elasticity as +177.3 against a computed +177.2; c04 printed CoW's within-bucket slope as -0.052 to -1.525, which WAS correct — the slope was computed inside the function and discarded before returning, and is now returned and interpolated. The first two are corrected in the companion paper; the third needed no change. The deposit carries the analysis code for both papers alongside the data: orders_v5.csv is the file to analyse, orders_v3.csv and orders_v4.csv are inputs to its rebuild. Order-level record of 480,572 executed trades across three order flow auctions on Ethereum mainnet: UniswapX (245,668), CoW Protocol (134,990), 1inch Fusion (59,992), plus 39,922 1inch plain-LOP limit orders reported separately (resting market-maker inventory). Those counts are the shipped orders_v5.csv; the companion benchmark panel markout_panel_v4.csv holds 480,554 priced orders. The Fusion/plain-LOP split is by decoded amount getter, not by bit 249 of makerTraits: the bit is necessary for a Fusion order but not sufficient, and of the 173,216 extension-bearing settlements 93.01% are verified Dutch auctions while 6.99% are not (4.04% 1inch FeeTaker, 2.63% no amount getter, 0.31% a Settlement getter whose AuctionDetails does not decode). Coverage: blocks 20,700,000-23,276,483, 2024-09-07 to 2025-09-02 UTC (360 days); ETH vs USDC/USDT/DAI/WBTC. Each order carries exact on-chain amounts from settlement events, block timestamps, the Binance one-second mid at the fill second, a same-block Uniswap V3 counterfactual across all four fee tiers, and amortised transaction gas. The archive also includes full-year Binance 1-second bars for ETHUSDC, ETHUSDT, ETHDAI, ETHBTC, BTCUSDT (~170M rows) so any benchmark can be recomputed independently. Papers: "Size Dominates, Gas Decides: Execution, Timing, and Flow in a Year of Order Flow Auctions" and "Filled Before It Decays: Identification and the Size Gradient in Order Flow Auctions on Ethereum Mainnet". All three venues also operate on other chains; this record covers the Ethereum mainnet deployment only. Version 2.0.1 adds four inputs the 2.0.0 changelog promised but did not ship (cow_solvers.csv, cow_orderbook.csv, cow_orderbook_deep.csv, uniswapx_orderbook.csv) and replaces that changelog, four of whose figures did not survive the final definition of the Fusion population; see CHANGELOG.md. All three venues are multi-chain and this record covers the Ethereum mainnet deployment only. See orders_v2_README.md inside ofa_orders_full.tar.xz for the column dictionary and structural caveats.