Originally Posted by Brian Ratliff

But this statement proves my point. You have zero information other than past data. You use that past data to fit a statistical model which you use to predict the future. Nowhere does information about failure physics, or manufacturing methods, or factory technician skill, or any other tangible, physical property of these frames is being used to create your prediction of the future.

I have a model too. It is not statistical though. My model says humans generally look for monetary advantage; quality control costs money, paying a kid 50cents/hour to sand and bondo a frame prior to paint is a negligible cost compared to tossing a frame or actively controlling quality up and down the line. In an environment where legal recourse is not a consumer option, the transaction does not occur face-to-face and the buyer never sees the product prior to sale, sellers will tend to take advantage because this is the direction the incentives lie.