Can I point you to my thread in
http://www.safeguardrobotics.com/default.aspx?tab=forum
under "How to... -> Using the CNC Brain with milling M/Cs" section where you will find my thoughts on isolation, earth loops and their relation to instability.
Initially, I would disconnect as much as possible in terms of I/O, encoders and axes you are not testing at that moment to reduce these problems. Swapping the drives round will prove if they are at fault or if there is a problem with the Brain. It's quite possible that one or more tracks are shorted on the Brain because the layout is very fine. I know of at least two boards that were faulty, one being mine. It would be good to test all of the axis outputs on the Brain.
You can alter the pulse width on the Axis tab but I doubt if this is the cause. I had to shorten this to the minimum because of the 1 micron resolution and high speeds I'm using but I doubt if you will come across the same problem.
Other than that, I don't think there's anything there that will affect it.
Are you getting random e-stop messages? These are primarily caused by earthing and isolation issues. The real pain here is the USB connection which grounds the Brain 0V. It's disastrous to have any other connection to earth. This happens as soon as you connect an encoder of linear scale! I know you are running open loop but I inadvertently shorted a 0V on a limit switch to ground and the system was completely unstable. I'm currently using full opto-isolation for ever connection including the USB and the system is pretty solid other than the software issues.
I hope that helps.
Roger