32 M9K Instanced as 16K RAM; After reset, the write the RAM fixed data , and always read when running.
But there is a strange issue that in 3 of 2000 pcs， after running several hours, one bit in fixed address is error.
This issue could repeat after re-programming .
Did anyone meet this issue?
Never had a bit error that didn't reveal as design error later on in my own (hobbyist) designs.
Are you sure your design has enough timing margin to sustain individual chip variation? What does TimeQuest say?