@NautilusTrader Treating time as an injected dependency is one of the cleanest ways to keep backtest and live behavior aligned. It also makes replay, timeout, and recovery tests deterministic—especially when exchange event time and local receive time are both preserved.
@QuantConnect Interesting result. The next evidence gate I’d look for is strict separation between signal construction and fill simulation: point-in-time volume features, realistic queue/slippage assumptions, and walk-forward stability after fees. Execution assumptions often dominate.