So, I finally got my Torus Pro Servo mill and Pathpilot situated and wanted to start making some parts. I load up a program that spots two rows of holes roughly 5 inches apart, along approximately 12" of travel. Next tool then comes and drills the through hole. Well I watched in horror as drill #2 comes down, makes a hellacious noise and I can see the drill bit bending while drilling. I stop and look at the the one and a half holes I drilled are ovaled. They were at least an 1/16 to an 1/8 off in both the x and the y axis. So I look at the code, and double check all the co-ordinates for holes are the same. So I do a bunch of testing, and find I can easily replicate the error, by zeroing out a dial indicator against a vise, raising Z. then rapiding from one end of the X axis to another and back, and then lowering Z and reading the offset. A single pass test on the X axis would randomly generate an error of .018 to .028, randomly in that range. The more passes I did, compounded the error. I found some great LinuxCNC tutorials for calibrating the ball screw, and checking backlash. ball screw setup was correct, my backlash measurements were ever so slightly off vs what is in the files being passed around by everyone for PathPilot on the TPro. One thing I discovered when doing all the testing though, is that if I used the Shuttle and manually stepped through testing it was fine. I wrote some G-code to streamline the testing, and it starts skewing again. So as a last resort, I took a look at the ini file, and backed the max velocity and acceleration off big time. Now it behaves perfectly, even with the rapids.

TL;DR
1. Massive positioning errors discovered and verified to be a machine (not g-code) issue.
2. Error goes away when maximum velocity and acceleration are lowered.
3. Concern that servos aren't servo'ing, or at least meeting my expectation of what servos are supposed to do.

Questions:
1.Is this down to PathPilot being configured for a stepper based system and does not look at the feedback loop?
2. Is the motion controller even in the loop at all? or does it just pass off a request for X number of steps to the servo drive, and the servo drive has the feedback loop and responsibility for keeping sync?

Looking for any help or guidance if anyone has anything to share. I'm a hobbyist, so if ~200ipm and slower acceleration is all this machine can do than I can cope, but I paid extra for servos because I thought positional feedback was an important element to making good parts. Side note, I understand that PP is not what the machine ships with and is way out of bounds from a support perspective from John@Novakon. Not a knock or gripe about the mill, this is entirely of my own doing and I accept that. I know that native LinuxCNC is totally capable of working with Servos, so if I have to go that route eventually, I will.

Bonus: Take the time to check backlash on all 3 axis and adjust the ini file accordingly. Each machine is a snowflake, My machine is essentially brand new and the backlash actually ended up being less than what the config is that folks are passing around. My X and Z were both .0005 off what was in the config. That error will compound with every change of direction. I'm not making anything that's going in to space, but make everything as good as your can.