A few dumb questions about loop closing

I have a mill upgrade project and it has some old analog servos that I'd like to make use of. All the analog servo drives that I can find take as input the encoders on the servos to "close the loop" for that particular axis.

The machine already has glass scales for a DRO that came with the original controller. So ideally what I'd like to do is feed the encoder values and the scale values back to LinuxCNC and close the loop there instead, so that it can coordinate each axis and can use the DRO output to handle backlash compensation, ball screw nonuniformity, etc.

My questions then are this:
* Is it a problem if I have a drive that "closes the loop" for its axis, as well as feed back the encoder values into LinuxCNC for it to do there as well? Will the error signal corrections become twice as big if I do that?
* If I do want to do that, is it just a matter of running a second wire from the encoder output on the motor to the controller (first wire would be encoder output to the drive)?
* Is it even worthwhile to use those scales for backlash compensation and whatnot? Does this work the way I've been thinking it will / as well as I've been thinking it will? How would it compare to getting a laser mapping of the ball screws for a compensation table?