The Arkansas story: "Glitches didn't hit everyone in primary", is the one that troubles me.

Assuming the same electronic voting machines, why would the same machine perform differently in the same circumstances?

They wouldn't.

It could be the use of Kelly temps to haul the voting machines to and from the precincts? Or even permanent election officials keeping them in their homes and offices?

The big story is the variation in performance under equal circumstances. That does not pass the smell test.

Black Box Voting (link here) has released an update which deals, among other things, with the issue of different results in the same circumstances:

"Two machines with completely identical software release numbers had different behavior with the same macro. Machine A just had a software crash and become unstable, while machine B produced an error message on the system log and contained the error while still resulting in loss of software functionalities. There were also other examples of different, but reproducible, software behaviors between machines with both modified and unmodified macros" (link here, bold added).

Every time I have seen this happen (assuming the version id numbers are correct, and that the software is in fact identical, as is the hardware) it shows that the stack or heap or some other critical area of memory is being corrupted somehow.