The only desperate way to do this would be take adv. of the fact that the DHT grids
are at 0 volts at idle (since you are using a cathode bias resistor and common filament
transformer). You could then apply a small, additive DC input at each grid - using a
potentiometer and resistor network. You could use a 9V battery, as that's as much
trim as you'd need.
Again, this is desperate, because obviously each DHT has a different transconductance, so the moment AC signal is applied to the grids, there'd be AC assymetry about 0V as
well as an "error" AC signal at the cathode bias resistors.... Only for idealistically matched
pairs, would there be a pure DC voltage at the cathode R and no AC signal at all...
If you go read some of Lynn Olson's stuff, this induced error voltage at the shared cathode can actually be dealt with through the "WE Bypass" topology. (See Lynn
Olson's Amity amp)
Most PP transformers are gapped to handle DC imbalances as you describe and perhaps
the solution to your problem is to try the WE bypass capacitor approach, or, need be,
get an output transformer that can take some idle imbalance current, and - as someone else mentioned - try to find tubes more closely matched in gm.
-- Jim