It isn't about mfg tolerances. Even the most precise manufacturing and machining processes leave the bearing surfaces with some microscopic imperfections. More importantly, the very machining processes (casting, forging, cutting, grinding, etc.) leave residual stresses in the substrate. It takes some time - and repeated heat cycles - for the residual stresses to work themselves out. That perfectly round cylinder bore may not stay perfect round as machined, once the residual stresses relieve themselves out... even if it is only ever slight so. This is why many engine builders often prefer to work with engine blocks that have good run time on them, instead of green components.
With any engine, the only thing that keep the mating surfaces from crashing into each other and creating accelerated wear is a thin film of lubricating oil. The integrity of that thin film lubrication depends on a lot of factors. One of the biggest is ENGINE LOAD. The higher the load, the more force/pressure is acting on that thin film, which leads to higher shear stress, which leads to higher localized temp in that thin oil film, which leads to localized reduction in oil viscosity and film strength, which leads to higher likelihood to localized breakdown of the thin film lubrication, which leads to accelerated wear.
Granted, the very process of "breaking in" involves some wear of the mating surfaces - smoothing over microscopic imperfections, and honing over stress-relieved microscopic undulations. The question is: whether it is better to do in a more controlled manner over a longer period of time at low engine load. Or... get it over quickly - and potentially rather violently - at high engine load. Which you should choose has become almost a matter of religion. Racers can't afford to do it slowly, so the latter is what they choose. To them, it is a better compromise. So if it works for Formula 1 racing teams with multi-million budget, is it the best for the regular Joe Schmoe who just bought a street bike?
