Well... the main bearing should have "the correct amount" of interference. Too little, and the main bearing can spin in the case (this can sometimes first happen when an engine is shut down for only a couple of minutes and restarted, as the cases will actually get *hotter* for the first couple of minutes that airflow over them is stopped). At that point, the case (i.e. bearing bore) is at its largest point.
Aluminum grows right at 12 millionths of an inch per inch per degree... and steel grows almost exactly half that. Many bearings in karting are 52mm OD (2.0471" in diameter). So for every degree over ambient, the bearing bore grows about 25 millionths and the main bearing grows about 12 millionths (let's call the bearing just growing half of the aluminum to make the math easy... 12 1/2 millionths). So for every 8°F over ambient, you lose .0001" (one ten thousandths) of an inch of bearing interference.
I'll cut to the chase: on a KT100 case, if it's the newer die-cast style case... anything under about .0012" interference at 70°F and you will eventually spin a bearing (may take certain weather conditions, or someone doing the aforementioned RE-starting of the engine when it's still hot, but there's a decent chance it will spin a bearing within a few races or so). On older permanent mold cases (early production), you can safely run maybe .0002" less bearing interference.
So... here is the flip side: very early Yamaha cases had as much as .0022" bearing interference. This "squeezes" the outside race of the bearing to a smaller diameter (which reduces the amount of internal clearance the bearing has... over its entire operating temperature).
Not only does bearing interference in the case need to be considered, but strength ("beefy-ness") of the case, as well as cooling capacity all need to be considered when setting up the correct bearing interference.
Most people do not have the means/equipment to re-machine bearing bores in cases, and this is why it became common to run C4 main bearings instead of C3. Manufacturers didn't want a bearing spinning in a case, so they just bumped up the interference a bit (or lets say: set it a bit higher than required). The solution was for karters to "adopt" C4 bearings as being de rigueur (fashionable). 99.9% of the karters never check bearing interference, and probably only the better or best engine builders check it.
The combination of interference, correct internal clearance in the bearing, measuring the quality of the bearing, running the correct lube, and a myriad of other things is what gives an engine not only performance, but longevity as well.
fwiw...
PM
PS: I'll have to politely disagree with MikeFlipp -- there is a lot more to bearings than simply "good" and "failed". A WHOLE lot more, as a matter of fact.