Johannes, are you using master branch for your development work?
There is an issue here with Plasmac that might be related to your nan problem. It does not affect V2.8
forum.linuxcnc.org/plasmac/40097-nan-voltage?start=0
So Plasmac uses some screen based variables in the GUI to calibrate a Mesa THCAD with values for scale and offset that are used to calculate Arc voltage. These values are sent to the plasmac.comp which parses it to a voltage.Some people are reporting sporadic nan errors for arc voltage when using their plasma cutter. As you can imagine this is not good when voltage is used as a process control variable to control torch height.
I wondered if the GUI variables are sent to HAL using NML messages. If that is the case, this must surely be related to your identified bug.
I'd appreciate your feedback.