68300

We have seen rare instances in our application when on a read of a 
flash memory part the part outputs 0xffff.  The part is not corrupt, 
a subsequent read will produce the correct data.  We have caught 
several of the incidents on a logic analyzer and have found that they 
always occur after a bus arbitration cycle.  The 68332 tri-states the 
address lines, the pull-ups start to pull them up, then the address 
lines are driven with a valid address and the flash is read.  The 
flash is used with the chip select always pulled, and WE and OE are 
toggled.  Does anyone know of a problem with flash when the CS is 
pulled and WE and OE are disabled, if the address lines are at 
invalid levels?  Also, is there a way to minimize the time between 
when the 68332 stops driving the address lines, and when it issues a 
bus grant?  (BG)