[Mondial]

Fields, try connecting a 100K resistor in series with the positive DVM probe and make the measurements again. This will filter any RF from the DVM input and hopefully give you a more reasonable reading. The series resistance will only affect the reading accuracy slightly, making it read about 1% lower with a 10 Megohm input DVM.

As Terry says, it appears you  have a gassy tube or an open grid resistor somewhere.