Freshness and Clocks in Multi-Venue Perpetual Market Data: Transport Capabilities, Staleness Practice and Timestamp Offsets

Cross-venue perpetual-market comparisons require data that are timely and timestamps whose meanings are understood. We audit public WebSocket capabilities at 22 venues and analyze 14,931 public funding-history records for BTC, ETH and SOL over 64 UTC dates in 2026. The new large-sample offset study aggregates to 4,977 venue-hour observations across Hyperliquid, Pacifica, Extended and edgeX. Median offsets after the nominal UTC hour are 37, 2,513, 849 and 0 ms; corresponding 95th percentiles are 88, 5,083, 1,593.15 and 0 ms. One Extended record at 1,096,849 ms is retained as an anomaly. A fixed calibration-period 99th-percentile-plus-1-ms rule attains later-period coverage of 99.35%, 100%, 96.61% and 100%, respectively, showing that historical tolerances need not transfer across venues or regimes. We also derive worst-case bounds for clock-difference exceedance shares when timestamps are missing. The results support venue-specific timestamp semantics, explicit anomaly reporting and partial identification rather than a universal staleness threshold.