Today I spent a few hours trying to replicate this issue again, but I was unsuccessful. I had CMOS battery removed overnight so I started with a clean slate. My HW test setup includes ISA ESS soundcard, PCI S3 Virge/DX VGA, 1x 256 MB 133 MHz CL2 SD-RAM and Celeron 366 MHz SL35S (I also tried Celeron 900 MHz SL633 and Pentium III 1.1 GHz SL5QW).
Step by step I was modifying BIOS settings to previous state and after every change I did a cold boot to pinpoint the problem. I think I can safely say that BIOS setting itself is not the problem. Originally (yesterday) I didn't do any changes to voltage, I only tried differences between pre-defined and manual CPU setting (always setting nominal values), but the problem was still there. Today I additionally tried to change voltage (undervolt), but the problem didn't appear. However yesterday I did change settings regarding SCSI/Promise UATA controller (pretty much every possible combination) because I was testing compatibility of this controller with SATA/IDE and SD/IDE adapters. This was quite problematic and involved aforementioned settings, resets at various stages and data corruption - it turned out that every adapter I have here is incompatible with the Promise controller. Some work only in Windows, in DOS they corrupt data, some don't work at all. But the controller works with IDE HDD, so it shouldn't be some HW problem of motherboard (apart from compatibility). Today I tried some of the settings combination for this controller, this time without these IDE adapters, and without any problems.
I can only guess that the testing yesterday got the motherboard's fail-safe feature into some kind of loop. When I got this motherboard a few weeks ago I also encountered this problem with interrupted boot (also extensively testing Promise controller), but I thought the fail-safe is triggered by Celeron 366 itself (fail-safe value = nominal value at 66 MHz FSB for this CPU) and using faster CPU will solve the problem. Unfortunately I don't remember what BIOS version I had flashed at that time (I was testing 1007.A, official 1008.004 and your modified version). Today I also wanted to test official BIOS in this regard, but since I am unable to reproduce the issue, it will have to wait 😀. Anyway thank you for your help/explanation and I hope I won't see this issue again.