Odd failure

I have a home built cnc router, running Linux Mint 20 and LinuxCNC 2.8.4 with Axis. G540 driver, XYZ plus A is slaved to X (there are two steppers on the X axis). I'm using DeskProto for CAM. This router has worked fine for several years.

I ran variations of some test gcode that puts holes at the corners of a 4" square. This worked fine the first two variations.
When I ran the third variation, it started with the hole in the upper left (no problem). When it moved toward the upper right, at about 1" it stopped completely. My g540 fault light was lit.
I cleared the fault by turning the power to the g540 off and then on. Resetting zero to the same location as before, and using the jog button to move the same way caused no problem. Moving the same way by manually executing gcode via MDI (same feed rate) caused no problem.
Going back to a previous variation (that worked before), I had the same problem. In this variation, the gcode started cutting upper left, then cut lower left, then moves toward lower right and at about 1"  stopped completely.

It looks like something broke (gcode that worked before no longer works, so it's not the gcode). I have not been able to detect the problem with either manual jog or manual gcode execution.

How do I troubleshoot this?

Thanks,
Steve

My first guess would be a broken wire...

This is a possibility, but odd. The problem only occurs during program execution. It's not broken in manual movements.

Might be power supply related (either drooping or insufficient output capacitance
to absorb BEMF when decelerating resulting in an overvoltage)
or maybe one of the G540 individual drives is failing

Wires were a good thought, I did find one sharp bend I did not like.
In the end, my problem was self induced. I tightened some places that I thought might help backlash--instead, I put too much friction in. The fault occurred during a rapid move as a result. It wasn't obvious to me since at the beginning it only occurred under program control; this changed to intermittent later.