all very interesting I'm sure but I do wish folk would stop wheeling this out as if it were gospel; it isn't. Sadly JB was wrong on several counts eg
1) you could have true brinelling and you would see the dents and not the raised material around the sides, because one would be a tiny fraction of the vertical dimension of the other so this 'observation' (that he didn't see raised material only indents) certainly doesn't "prove" fretting. Indented headsets can occur in just a few miles; again not consistent with fretting.
2) Brandt's steerer was absurdly long (over 10" between races IIRC), only 1" diameter and the frame was built light, so the steerer trim wouldn't have favoured stiffness, shall we say. There would have been significant deflections in the steerer which were not considered. In my own bikes I have one frame with a head tube about 3/4" longer than the others and the steerer is noticeably more flexible.
3)When the steerer flexes the load isn't shared between balls in a headset with standard raceways, so the whole load can indeed be seen by three or four balls only. Hence indenting via gross yield (brinelling).
All of which means that what other people were trying to tell Brandt was correct; his headset had an anomalous setup and his experiences were likewise anomalous. Red herrings such as hammering head races together and claiming that this is representative of service loads somehow just don't help either.
JB changed to a 1-1/8" steerer (less likely to flex) and a cartridge bearing headset (which better accommodates flexing). However the experience of anyone who uses frames with medium length head tubes and controls the preload on good quality (and well installed) 1" headset bearings correctly has always been that the headset is likely to outlast the rest of the bike.
FWIW I have seen numerous cartridge bearings which failed, and sometimes they damaged the seatings and took them out as well. IMHO if the bearings are not to move around too much there needs to be quite a lot of preload and the service loads are closer to the maximum load the bearings can sustain (because all things being equal, there are fewer, smaller balls in a cartridge bearing headset than in a traditional one. And whilst cartridge bearings can articulate on their seatings it isn't a very good arrangement if they are forced to do so on a continuous basis.
FWIW if you have a cartridge bearing with 36x45 writ upon it, it means one shoulder angle is 36degrees and the other is 45 degrees.
cheers