the reason loose balls work well to replace clipped balls is severalfold
1) there are usually appreciably more balls without a clip present (except in a few cases e.g. some campag headsets) so the bearing is inherently stronger
2) the loose balls are spaced differently to the clipped balls / 'indexing marks'
3) if you want to do a proper job, remove one of the lower races and replace it in a different position, so that none of the marks align exactly when the steering is in in the straight ahead position, e.g. turn the crown race by half a mark or something.
If you do all these things you might get one or two balls sat exactly over an extant mark; the rest however will bear against unworn parts of the races and will share the load. The net result is usually a perfectly acceptable headset. My view is that this is nearly always worth a go, because it can be done quickly and cheaply, without the usual pitfalls of exactly identifying the specification and tolerance of the original headset, buying the same, then removing the old one and fitting the new one without chipping the paint etc. etc etc.
It is a while since I read Brandt's explanation/diatribe but I didn't entirely agree with it. If his argument is correct then this would be an excellent reason for using a grease in headsets that contains high film strength synthetic oils, solid lubricants, corrosion inhibitors and EP additives. Greases meant for low-speed/exposed industrial bearing applications are made this way, so I suggest that greases of that sort are used in bicycle headsets; it certainly won't do any harm.
BTW his headsets may well have marked themselves up in the fore-aft direction (sort of) as he described but when I saw his bike I understood why immediately; he used an extremely long 1" steel steerer in a light gauge; the flex in this would have allowed the lower races to articulate through a much larger angle than normal ( I estimate up to three times more than in those steerers that most folk use). Once the articulation reaches a certain point the balls cannot share the load any more and some wear is inevitable.
My experience is that in steerers of 'ordinary length' (i.e. to fit into head tubes 6-8" length) such marks do not appear in good quality headsets (fitted with loose balls), even after many tens of thousands of miles, provided they are lubricated and adjusted correctly.
FWIW most headsets that go bad don't die of natural causes, they are murdered, i.e. they die because they are not installed or adjusted correctly. You would be amazed how many people tighten the top nut on an A-head too much, or finger tighten the adjusting race on a threaded headset before bringing the locknut down on it, crushing the life out of the bearings....
The correct method for adjusting a threaded headset is to set the adjustment so that it is loose when the locknut is tight, then back the locknut off, adjust the threaded race a little inwards, retighten the locknut, and check for play. Repeat until there is no play. If you adjusted 1/10th of a turn at a time, this is up to about 100 microns (i.e. a country mile in bearing terms) preload so there is probably a better setting somewhere between the last two positions you tried. With good quality bearings you should be shooting for about 10-20 microns preload, i.e. a few degrees or one or two 1/100ths of a turn on the adjusting race past the point where (with the locknut tight) there is still detectable free play.
As someone pointed out above if you have a very flexy steerer, some of the articulation may be (again imperfectly) taken up by the angled bearing seatings in a cartridge bearing headset. However this is (IMHO, along with easy cartridge replacement) pretty much their only positive aspect; size for size they tend to have fewer, smaller balls in them (so have a smaller load capacity), yet require a higher preload so as not to rattle over bumps etc. I have seen plenty of cartridges that started to 'index', mostly because they were badly adjusted.
cheers
