Thanks for the feedback guys.

I found a reference so I could start learning G-code. It looks like the G-code never turned off the torch, so I'm guessing it was due to the noise on the interface from the motors when they cut the corners and doubled speed and made quick direction changes.

I didn't put a ferrite on the BOB -> THC cable yet. That's probably the first step. If that doesn't do it, I guess I can delay the torch off in software (probably start with 250 ms and take it up from there). I'd prefer to address the noise, but I think I could end up cutting/wasting a lot of steel getting scope captures to figure out what's going on.

I think that the cut rate was way too slow. I had been trying to find a good chart to give me a starting point for cutting speeds and amperage settings based on thickness. I finally found a decent starting point on Eagle Plasma's web page:
http://eagleplasma.com/4x4system.html

It recommends 240 ipm @ 30 amps for 16 ga. I cut the 18 ga at 10 ipm @ 35 amps.