I've run into an odd problem in an ATxmega256A3 design. On two (so far) of 10 prototype boards, after some period of proper operation, the boards appear to restart when an attempt is made to write EEPROM. The errors start out as intermittent, but over the course of a few hours (I'm calibrating things and the coefficients get stored in EEPROM) the reset rate goes to 100%. I've traced the fault to the execution of the low level EEPROM write provided by Atmel's eeprom_driver.c file from app note AVR1315. I can't say exactly where things blow up, but it's during the write.
Watchdog timer is off and I haven't yet checked the reset flags to see if there's a hint there. It may not be a reset, but rather the code walking off into the weeds and eventually wrapping around to zero. Regardless of the cause, it doesn't happen on every board and it doesn't happen using the same routines on other xmega variants I've used in the past.
Anybody got a clue?