I agree with those who say it is the order of connection, or the resistance right at the point of connection. It cannot be anything else. The big problem is why the resistance would be lower at one connection point. It may be that something is distributing the connection at the grounded neutral or "cold" terminal causing a larger cross sectional area (and less resistance) for that termination, or something is concentrating current in a smaller cross sectional area at the "hot" terminal causing the same current through a smaller area (more I^2 R heating from more resistance around that termination). It cannot be anything else.
This reminds me of a clock. My 120V electric motor ship clock was subjected to a lightning surge that shorted near one end of the winding to the motor frame. This made it sensitive to stray capacitance or leakage to ground. If I installed the plug one way, it ran forward. If I installed the line plug the other way, it ran backwards. I had an unbelievable amount of fun with that clock, until I discovered the clock face was at 90 volts or 30 volts to ground, depending on insertion. The pickle, however, cannot be like the clock because impedance levels are far different. It cannot be sensitive to displacement currents via capacitance like the high impedance clock motor.