Sawtooth PID error

Hello all. I'm tring to get rid of a small steady state following error on the X axis of an XYYZ cartesian plasma cutter [see first screenshot].

Any amount of P produces a sawtooth error pattern [see screenshots 2, 3, 4]. I can't find any combination of P, I, D or FF1 that even starts to get rid of this jumpyness.

Has anyone experienced this before?

I'm beginning to think that the error is somewhere else in my configuration and I'm not seeing it, so I've also uploaded my hal and ini files.
Thanks in advance for any help you can offer.

What could cause cyclic perturbation like that? Would seem like it might be a mechanical issue, once per revolution of a messed up belt, pulley, bearing? Or maybe a latency or electrical interference spike every 1.66secons? If you move at a different velocity, does the frequency of the error pattern change? What does the velocity feedback plot look like?