Comment by LorenPechtel

The shuttle people never looked.

That there was blowby where there should be none was known, but nobody dug into why. There was no determination that it wasn't enough to be a problem, just an observation that it hadn't blown the booster up yet. For something with a wide spread in the data points, no way to model the maximum expected values.

Once they got serious about looking it didn't take much to reproduce the problem. They built a single joint, mostly filler inside, fuel to model the real thing during ignition. Maybe worked, maybe spot-welded, maybe complete failure. The colder the more likely to fail.