No idea what the later bus standards (PCI, etc.) do about it, but hard freezing is a traditional 'panic' mode in the IBM-PC when /NMI is pulled on the ISA bus. This is usually reserved for "shit's broke" signals, and the usual BIOS ISR (interrupt service routine) is an infinite loop (JMP +00 or something like that). So interrupts remain disabled, the processor doesn't read any memory but the BIOS that it's looping on, and peripherals are left alone. Since the keyboard ISR isn't used, CTRL+ALT+DEL will not clear this fault, and a hard reset is required.
I would be surprised if this sort of signal and code path is still present, but it could be that something vaguely similar is happening. The traditional "panic" of a 386+ user-kernel system (typical of *nix and Windows NT+) is an escalating processor fault (e.g., page fault, general protection fault (GPF)) that causes the kernel code itself to crash, triggering a reset (and on Windows, a BSoD, at least briefly before the reset vector is executed).
Could very well be that the driver is so poorly written that it doesn't play nicely with any of the systems integrated on the SBC. I have no idea what driver or kernel programming is like in Windows, so... could very well be, they don't know either..
Tim