separte dc ground in tunable filter

For tunable filter with varactors, is it ok to use a separate dc ground (for biasing varactor) than connecting it directly to RF ground?

As a general rule there needs to be a definite current return path, when a device is either powered or controlled by current. In case of a complex circuit, it's not always easy to be certain that it provides this return path.
I suppose this does not answer your question directly. If a path to ground is not available, or if it changes between low and high impedance, then you might consider creating a path to the positive supply. Etc.

Provide bypass capacitors to RF ground next to each varactor.

yes, you can separate the two. Lets say you have three shunt mounted diodes hanging off of some sort of transmission line. You can apply DC Ground to that transmission line. The other end of each varactor diode can go directly to a 100 pF chip capacitor, maybe even be physically on top of a ceramic chip capacitor. Then the tuning voltage would be injected where the varactor touches the that capacitor