Erik Malund wrote:

NOPE, the fact that code is "burned" into the EE/Flash does not guarantee that it is there tomorrow. The famous "fast programming method" that is still popping up now and then was based on "the programming pulses does not 'need' to be as long as the datasheet states, the thing program very well with 1/5 the width" had the little drawback that the programs vaporized in a few days/weeks/months

Well, I don't disagree and this might be the case. But do you have any references?

I am getting curious on this topic, since I never heard about that phenomena. I also contacted Atmel, but they gave no satisfactory answer.

Please enlighten