This is not strictly Slackware-related, but this problem appears on my Slackware HTPC, so I thought that someone might have experienced it before.

I have an Intel NUC machine, with an IT8625 SuperIO watchdog chip, in addition to an Intel CPU's built-in TCO wathchdog.

The IT8625 is supposedly supported by Linux Kernel, using the it87_wdt module.
Moreover, it is possible to initialise it right from the EFI boot sequence, setting the timeout in the BIOS Setup.

However, there are issues:

1. If I do not load the it87_wdt module, the system reboots shortly after boot, as expected. If I _do_ load the it87_wdt module, however, the timeout stops, and I found no way to trigger it. Echoing anything into /dev/watchdog does not seem to do anything, as well as loading/unloading of watchdogd (from SBo, by yours truly).

2. Sometimes the system hangs on loading initrd, at GRUB. Like, if I interrupt the GRUB timeout timer, by, say, selecting a non-default boot option, but not pressing Enter, the system is rebooted. However, if I don't touch the system, sometimes, maybe abount once in 10 boots, the system hangs at
and it hangs at this stage forever, and is never rebooted despite watchgod timeout being set in the BIOS setup.

If there is no initrd, this problem does not manifest itself, so I temporarily avoided it by just removing initrd altogether. However, this is imperfect, because using TSC_DEADLINE on this CPU requires a microcode update, which is usually done from initrd, and generally having an initrd is nice.

I haven't experimented with appending an initrd directly to the kernel, but this requires a custom kernel, so is also imperfect.

Any anyone suggesst anything?
It really seems like a buggy chip, but I believe this forum has a lot of people used to dealing with buggy hardware.

I believe your diagnosis of "buggy chip" may be spot on: https://github.com/lm-sensors/lm-sensors/issues/186