[Amps] Peculiarities- perhaps slightly OT

[Peter Chadwick]

So I'm trying to get set up on 160 to get 3B7C. The antenna is the tower, fed as folded unipole, via an  L network, remotely tuned. The variable L is from a old RCA ET4336, that ran a pair of 813 in Class C at about 800 watts: the variable C is a ceramic bodied 500pF vacuum variable rated at 7500 volts. In shunt, switched in with a relay on 160, is  220pF 10kVA ceramic padder. The tuner is in a weatherproof box on the side of the main antenna tuner shed, and has a 15 watt heater in it. Max temp is 48 deg C: min is 2 deg C - there's a Six's max/min thermometer in the box..(The same antenna is used on 80, and did work 3B7C on about the third call.) The power level is around 100 watts because the amplifier has been undergoing rebuild since last September - illness hasn't helped.
So the first problem was an intermittent SWR that I eventually traced to the 160m relay not always pulling in because the ULN2803A 500mA octal driver had decided to sit with 9.5 volts across it when it was supposedly saturated - and only sinking 150mA, when it's rated for 500. No spares to hand for some reason, but a 1 k resistor and TIP115 PNP Darlington provided sufficent of a bodge job to ensure the relay got pulled in. Next problem is that the tuner tunes, the SWR comes down and after about 15 seconds of 100 watts, SWR goes racing up. Suspect the ceramic 10kVA rated padder capacitor, so exchange it. Same thing. Substitute an air variable for the ceramic padder, on the basis that the ceramic has too many kVA (unlikely) and is heating and drifting. Still get SWR drift. Unhook the vacuum variable, use the ceramic padder and air variable, no drift. 
Now this seems strange. I've not heard anyone here mentioning capacitance drift in a vacuum variable. Reconnect the system as original , and lo and behold, no drift. The connection to the vacuum variable is a ring with a Philips screw and 6-32 nut and was clamped down hard. The grounded end of the Vacuum variable is a spilt clamp of 3/8 thick aluminium, done up tight with a hex head 1/4-20 Allen screw. And I mean tight!
Because the box is heated, it would seem a valid conclusion that there's minimal condensation and corrosion. The connections are tightly clamped mechanically, so what causes the problem? And will it reappear?  For that matter, will I manage to get 3B7C on 160? Time will tell.........but this variability seems very strange to me.
Any ideas, guys?
73 
Peter G3RZP