DMF 

(Not talking about headsets. I grant you that one.)

With all due respect, you're not reasoning. You're extrapolating from subjective experience. (Nor is your experience universal. Mr. Dopolina has related that his experience with hubs is different.)

Further, your experience is of an extremely limited subset of the total functional life of the bearing — typically only the first few rotations.

However, the scientific method is partly based on experimentation, so experience has a place. We just have to be careful how we use it. Try this.

I challenge anyone to feel or listen to a functional hub or BB, loaded or unloaded, after 1000 mi. of use and tell whether it has loose balls or a retainer. You should not know what type of bike or hub/BB it is. For statistical significance*, you have to be right (or wrong!) about 9 times out of 10.


*And technically, the result must be adjusted by the actual occurrence of each type in the tested population. E.g. if 90% of the samples are of one type, then a random response yields 90% accuracy. So 99% accuracy would be required.