Impregnation is a decision, not a default

Impregnating the borderline leakers and scrapping the rest. Twenty two of thirty nine came back and passed.

Trying to impregnate every failure. The severe ones came back and failed again, so the labour was spent twice.

Sorting by leak rate first and only impregnating the slow leakers turned an eleven hour rework into a six hour one on the following run.