Once with a very early Atmel chip I reported a problem when the revision was in the neighbourhood of 'g'. Atmel vehemently denied that the problem existed, but lo and behold, about a year later it came out in an errata and was fixed in a revision about 's'.
I totally understand that there is no way in hades that each and evey permutation of a chip can be verified by the makers so asking for a first run perfect chip is not reasonable. What we CAN ask for, however is errata being released the very moment a problem is known. This also known as basic honesty.

Erik