Unexpected Realtime Delay on task 0....

I have...

Disabled all there is to disable in the bios.
I've set Isolcpus=1,2,3 (quad core pentium 3700).
C=1 (I think 1, whichever runs full blast all the time)
Disabled sound
Disabled everything I'm not using.
No Wifi
Secure Boot is unchecked in bios
I've disabled SMI via SMICTRL

I per the guy with the goat avatar, PCW, the histogram looks good.

uname -a returns:

Linux printnc 4.19.0-17-rt-amd64 #1 SMP PREEMPT RT Debian 4.19.194-3 (2021-07-18) x86_64 GNU/Linux

i suppose I could find an ethernet USB dongle and disable the onboard Intel ethernet. this is a single board computer (NUC5PPYB I think) so no PCI slots or way to add otherwise.

Strongly considering a raspberry pi

This has no case. it's in my electronics box. Will try 2.7. It's got only one memory module.

I get errors randomly, often just when it starts up, so I'd be shocked if it was a heat issue. It had been running for a couple hours. I rebooted, entered bios and the memory temp was 31C and motherboard was 36C

My gut says it's an ethernet issue.

Hanging on boot from USB. I think its expecting the ethernet adapter to be eth0 but it's enps02 or something like that. this is an intel board and intel ethernet adapter (onboard).

Not able to tinker at the moment... work.

Okay so I hesitate to post this... I don't want to jinx anything but I might have... just maybe... solved it.

I did some digging on the network adapter that's on the board. It is a RTL8111/8168/8411 Ethernet Controller. The driver that's part of Linux is r8169. I read that some folks had network problems with it. Connection dropped and etc. I confirmed I had that same version by running lsmod

I then confirmed I had the earlier version too with sudo apt show r8168-dkms . So I installed it with: sudo apt install r8168-dkms

It all went swimmingly til my /boot/ partition was full and the "initramfs" got corrupted. fortunately I sat that I had to fix it before rebooting so i went down that rabbit hole, which seems par for the course in Linux.

ANYWAY. I ran a couple GCode programs and have been watching it for a good while and I think/hope/wish I've got it solved.

Fingers will remain crossed.

Me again. sheesh. So I've been bitten by the "watchdog" twice now. Usually just after starting LinuxCNC.

Watchdog timer is 5000000 (5 million)

I notice now after reverting to the earlier driver that my ping times are worse pinging 10.10.10.10 (7i96 board). Getting about 1.18ms. Seems rock solid though, nothing lower than 1.17ms and nothing higher than 1.20ms.

halcmd show param *.tmax shows:

4338760 for ...read.tmax
3880520 for ... write.tmax
5483520 for ...servo-thread.tmax

not sure what units are on those. What can I do? tame the watchdog? make his timer longer?