Soda-san...
This is a long-shot, but is it possible that R4030_SYS_IT_STAT is in a region
of memory being cached???  That would explain the fact that reading this
address isn't clearing the INT4 interrupt pending.  I was thinking that maybe
machines with L2 cache (like the Magnum 4000SC) were being inadvertently setup
to L2-cache addresses which it shouldn't.  This would also explain why the same
code works fine on the 4000PC.  Just a theory...
What do you think?

-Mark