Simple logic:
If it packs itself up in short mileage again-and-again, then it's 99% sure it's wrongly shimmed. That approx 1% is left for radial play, that is very rare - the main shaft in FD is glued together from 2 parts, so the glue can crack and it can develop radial play.
9 cases out of 10 you do not need to re-shim, since the new bearing IS the same width, at least if it has the same part number (usually it's FAG part). But as sayed - if it blows in short mileage then very probably it is wrong shimming.
I'd say: if your bearing goes over 50K miles w/o any problems, then DO NOT re-shim, use the old shims, since it's normal wear. If less than that - RE-CHECK the shimming and re-shim. An exception would be if you do freqent hardcore offroad, extreme hot conditions, ride in third-world potholed roads with fully loaded bike etc. abusive use for the drive-train and axles.
I replaced mine just as a perventive maintenance @ 65K miles, because the seal started to slowly leak and I had only 1 seal + 1 bearing set, so if the bearing had gone later (then usually the seal also goes, even if it's new) then I wouldn't had the seal in the middle of nowhere in South-American patagonia. Judging from the fine metal dust on the magnet (no big pieces yet) I'd say the bearing would had done at least 10-20K miles more - and that's mostly 2 up full geared bike (around 450 kilograms!) + bad third-world roads. So they do last if they're correctly shimmed and oils changed frequently (I do @ every 6K miles if possible). In extremely hot, fully loaded bike and hard conditions I recommend to run xW-140 in the FD box, everywhere else standard xW-80/90 works just fine.
And NOTE that you have to check the correct shimming when the FD box is in the working temperature (around 70-80-ish Celsius). Reading around the net lot of guys try to do it in cold and complain about blowing up too often - when cold you get plain wrong results on measuring the freeplay - I tryed to test it and I was amazed how much the freeplay varyes in FD with temperature.