Understanding how train drivers’ short-horizon control varies across operational contexts is critical for proactive railway Human Reliability Analysis (HRA). Engineering models often leave momentary human-performance variation unmodelled, while neuroergonomic evidence is difficult to transfer operationally. This study addresses this gap using Wickens’ Multiple Resource Theory (MRT) and within-run fixed-effects regression (MRT-FE) on 8291 runs from a British main line. Operational cues are conceptualised as dynamic Performance Shaping Factors (PSFs), and local acceleration changes serve as a behavioural proxy for short-horizon control stability. A speed-limit headroom diagnostic shows that high-magnitude adjustments concentrate under reduced headroom and compound infrastructure conditions. The diagnostic supports the proxy’s safety relevance; future validation against failures and violations requires verified safety-event records. Results indicate that short-horizon control adjustments are associated with compound PSFs, with non-additive patterns across MRT-guided visual–spatial overlaps, rule-coupled station–signal contexts, and cumulative-distance interactions. Representative patterns include deceleration-like shifts at tight-curvature tunnel exits and context-dependent reversals of cumulative-distance effects across geometry and infrastructure conditions. These findings support a shift from static thresholds towards dynamic, context-specific interventions accounting for interacting operational cues and infrastructure complexity. The MRT-FE framework quantifies and interprets associations between compound operational contexts and short-horizon control stability, thereby informing proactive HRA.