Here’ my last guess: Sloppy redraw shows what I think the problem is. With Switch A closed, and points on left opens, the counter EMF and condenser discharge flows to the right side and to ground via right side points and coil primary winding (orange track).
Closing Switch B puts it in parallel with the above track.
Conclusion: There is a surprising whallop going through that circuit. +/_ several hundred volts, actually. I think you have too much resistance in your kill circuit, and it is not dissipating all of the energy, leaving enough to create some sort of weak spark in the running cylinder. Yeah, I know, you Ohmed it all out. Probably with low voltage and less than a MA. Prime suspect: Switch A rating. Switch B in parallel provides a higher total rating. Second suspect: the wires size or broken strands.
OK, ball is back in your court.