I would think the locomotive would require more current to move at the lower voltage setting, so I'm not sure I agree with the analysis that this has to be an over current issue only at 18 volts.
I tested a Pulmore motored locomotive with four lighted 15" passenger cars at 14, 16, and 18 volts track power. The start-up and running current for the same observed speed was more at the lower voltages in every case. This was using a pure sine wave transformer and a Fluke clamp-on ammeter.
I think something more than simply over current is in play here. I could certainly be wrong, but I wouldn't bet all my money on lower current at lower track voltages, especially during start-up and attempting to maintain the same speeds.
I'd still like to see the noise suppression cap tried. I'm not saying it'll solve the problem, but the process of debugging is simply removing possibilities until only one remains.
