Green operating: "A" = 2.0V across LED. Current = 3.2 milliamps; calculated by the remaining 1.5V across the 470 ohm (current = 1.5V / 470 ohms).
Red operating: "A" = 1.7V across LED, Current = 3.8 milliamps.
Green failed: "A" = 0.3V across LED. Current = 6.8 milliamps.
Green and red voltages and currents for the operating condition seem OK.
The failed Green is curious- it draws more current than when operating...yet does so with only 0.3V across it. I can't explain the why but it seems like the spark induces a "latch-up" condition which can be reset. What's curious is that both Greens fail together - my guess is that one fails and drags down the other so to speak. To that end, since it appears this is easily repeatable, I'd be curious to see what happens if you remove one or the other Green LED wire, induce the failure in the remaining Green LED, then re-attach the 2nd Green wire. Obviously you can go wild dreaming up experiments to narrow it down, swapping parts, etc. etc. As for a remedy, I'd put a small capacitor, one each across each green LED right at the signal head. If that's inconvenient, try just at the terminal block where both green LEDs are connected. Something like 0.1 uF (microfarads) or really whatever you have lying around in that ballpark is a good starting point.
Very interesting indeed.
