Experiment Note
An initial communication-noise sensitivity sweep was found to be invalid because the intended configuration keys did not actually change the experimental conditions. The output varied in presentation, but the requested noise setting was not being propagated through the simulator as intended.
The sweep was excluded from the evidence base rather than reported. The manuscript explicitly recorded this decision in the threats-to-validity section. The final diagnostic package relied on experiments whose settings were verified: held-out attacks, platoon-size sweeps, headway sweeps, delay stress, latency, OOD sanity, representative trajectories, and SUMO validation.
Technical Essay
This was one of the most important quality-control moments in the project. A sensitivity plot can look plausible even when its control variable is not changing. Configuration-driven research code therefore needs explicit logging of the resolved parameter values, not only the requested command-line arguments.
The experiment was not quietly repaired and rerun merely to preserve a desired narrative. It was removed because the attached manuscript deadline was close and the remaining verified experiments already supported the scoped claim. Future repository work should add configuration assertions and emit a per-run manifest containing the effective simulator settings.
SUMO VALIDATION AND SYSTEM-LEVEL ANALYSIS