>> It's usually not the hardware, it's something that overflows the
>> interrupt stack,
> Sure.  I don't think that broken hardware can overflow the interrupt
> stack.

Sure it can.  Consider what happens if the hardware that raises IPL to
match that of the incoming interrupt ends up zeroing the LSB of the new
IPL value: interrupts coming in at an odd priority end up not masking
themselves when tripped.  If any of them are level-triggered rather
than edge-triggered, boom, instant interrupt stack overflow.  Even
edge-triggered, if they happen faster than they get handled.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse@rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B