two inserts fit the same FR10 inspection base

It supports it well enough to fool me at a glance; we compared them on the bench and the difference is underneath, where one misses a locating face.

Then I wouldn't make recognition depend on how nicely the cover sits. You can get a very convincing wrong answer that way. What identifies each insert when the cover is already in it?

Put both beside the base before buying sensors, because if the only identifying mark gets hidden by the part you'll have paid to automate the same guessing game.

Nothing visible with the cover loaded, Rosa, and quality found that the wrong insert lets us check against the wrong face without anything looking obviously crooked.

That is the bit to show whoever changes the tooling. Not just two inserts with similar holes, but the same cover looking seated against a different reference. I'd want the identification visible, and a way to stop that wrong combination seating as if it belonged.

Labels are still worth doing, Maya, just don't give them the job of a locating feature; somebody has to know which loose insert they're carrying over.

Toolmaker has proposed a different locating arrangement for each family, with orientation built into it, plus an exposed identity tab that stays readable with a cover in place; we're also changing the recipe sheet to show the actual insert instead of only the cover drawing.

The altered inserts are back and the wrong-family and reversed combinations won't seat on their bases in our bench check, which is a relief, but we haven't checked the recipe selection part yet.

Who is doing that last check with someone other than you choosing the recipe? I'd include a correct insert with the wrong recipe already selected. That is the combination your new mounting still allows.