1) My _hunch_ is that if the GFCI is tripping, then there is likely a real fault to ground somewhere. While you test these solenoid valves using a high voltage, it is possible that the fault only exists while the solenoid is moving or energized; testing for this would require quite a bit more effort than simply using a hipot.
2) As the OP suggests, a GFCI does not trip on a difference in current. It is _supposed_ to trip on a difference in current, but the method used is to run all of the circuit conductors through a sense coil, and the GFCI actually trips if sufficient net current is measured on this sense coil. If the sense coil is not perfectly balanced, then a sufficiently large _balanced_ current can cause the GFCI to trip. I doubt that this is the issue, but it is possible that the inrush current on the solenoid is causing this sort of trip.
3) On langjahr's comment about the current build up in a solenoid, a slight correction is in order. 'Inrush' is most certainly possible on AC solenoids, similar to transformer inrush. The R/L time constant of a solenoid for AC use is very short, and the current is limited by the AC impedance, not the DC resistance. The AC impedance assumes that the core is not saturated, and that the flux is smoothly alternating from a negative peak to a positive peak. But at startup, the core may have no flux in it, but the AC cycle demands a full + to - flux change. This will cause the core to saturate, significantly lowering the inductance. During this instant, very high current will flow. So the current starts low, and builds up over the course of 1/2 cycle to much greater than normal, and then dies down to the steady state case.
4) Langjahr's suggestion to look for capacitance to ground is another plausible plan for investigation. With larger motors on VFDs, I've seen 10's of ma of current to ground, with no faults and no insulation failures.
5) You might consider using a logging oscilloscope connected to a clamp on current sensor, and metering around your system to locate any 'zero sequence' or 'ground fault' current flows. Basically you would use the scope and the current sensor in the same fashion as the sense coil in the GFCI, to see what causes current spikes that look like ground faults.
-Jon
