I get MCU 'mcu' shutdown: Timer too close a lot, somewhat randomly once in a few printing days, but sometimes every time it’s at the same-ish place in the executed gcode of the print; the last two examples in the log are like that. I don’t have anything actively running on the SBC except klipper and tail -F klippy.log over ssh, plain Armbian on Cubieboard. No cron jobs. Reducing microstepping didn’t help.
You are using a modified Klipper version both in terms of the host version and the firmware part. If related to your issue is unknown and nobody here will follow up on such a version
So you are saying that you have already excluded all the reasons listed, e.g. SD card that is beginning to die, unstable voltages, additional USB devices, additional processes going rampant etc etc? If so, then maybe related to your modified version
This command is aggressively trying to get access to the file (-F) and returns every change. In the context of timing issues on the host, it is at least not what I would do. YMMV
I think you don’t understand how inotify works, and regardless, we can both see that it doesn’t impact CPU consumption much, if at all measurable. Besides, /tmp lives on tmpfs, so it can’t even introduce iowaits.