i have not read the almarg posting to which you are referring but the failure mode to which you are referring is highly unlikely in practice because semiconductor manufacturers typically manufacture devices to iso 9001 standards. so they do burn in testing to screen out the kinds of failures that you describe (which are typically "infant mortality" type failures) [i briefly described burn in testing in another posting but burn in testing by electronic manufacturers is different from "burn in" testing as the term is used by many audiophiles]. furthermore, semiconductors are tested over a temperature range that is wider than they are likely to encounter in actual use.
03-05-11: Hifihvn
The link to a post Almarg pointed out covers the way it is IMO. Some electronic components (semiconductors) do not like the big thermal swings, and may fail sooner. Others may settle, and break in to the on always operation since they stay hot all the time. With this case, those semiconductors may break when they shrink during cool down, since they have never been allowed to contract during a cool down, and expand during warm up, since this is what they have become accustomed to. Think of it as a super miniature bridge without expansion joints, for thermal change, while others may handle the temp swing like a bridge. That one poster is an engineer(Almarg link) that specializes in this. And like he says, it varies from component, to component. I'd have to agree with that. It makes sense.
but failures can occur at various stages. the semiconductors are assembled onto boards and the boards are assembled into systems. burn in testing would need to be done to screen for infant mortality at each stage. a given audio equipment maker may not use such quality control processes (it does after all increase costs because some products will fail) but bryston does do burn in testing before shipping products (according to the information on their website).