OK - lets take a step back.
The OP's original question is how to handle sorting of a supplier's material given that the supplier's process has shifted AND the original tolerances were derived using statistical tolerancing and not min/max stack up tolerances.
This is truly a rock and a hard place for the OP.
Statistical tolerancing requires that the component features are Normally distributed. This allows for larger acceptance tolernces than a straightforward stackup of fits as it would be 'rare' for all components to be at their max - or min - to create a fit problem. When a process goes out of control - and they all do - you get a bolus of product nearer or at one of the spec limits. This can create more stack-up fit problems than expected. When this happens it is best to sort out the bolus to minimize cost (better to cull it out at incoming than during assembly - or worse yet to experience increased wear failures in the field). This poses several issues:
- Where to draw the line? this could be done statistically but would require knowledge of the new distribution AND confirmation of the mating part(s) distributions as well as the necessarry calculations. It could be mathematically complicated or as simple as lining up the distibutions in a graph and drawing inspection cutoff at the intersection - my choice as it's simple for everyone to understand and doesn't rely on proper statistical choices. You could get really conservative and guardband the cutoff to accomodate measurement error as well)
- How to handle the supplier? this is typically dictated by the contract. If the contract has provisions for distribution, Ppk or maintaining statistical control, the choice to charge the supplier is easy. If the contract is silent on the supplier's responsibitility for maintaining the assumptions/requirements for statistical tolerancing, then the customer should eat the cost of the sort
All other discussions concern a horse that has left the barn...or our own personal soap boxes.