Which brackets belong in our FR10 correction allowance?

Two hole-spacing failures and one upright-angle failure. Each has a later pass, with no active correction time.

Keep those feature histories separate. Do the later passes use the same drawing revision and released checking method as the original failures? That needs confirming before the examples describe successful correction.

Quality and fabrication selecting the group. Revision D on the old checks; inspection is confirming the support methods.

Who writes down the first released result before the correction starts?

Inspector at the released check. Fabricator records active correction; inspector records the return result. Waiting stays separate.

That sounds less like one magic rework time already.

Show the estimator that distinction now. You can remove an unsupported universal allowance while the actual observations are still pending; the replacement should be visibly provisional, not silently zero.

And retain the selected-failure examples as examples. Their lack of a denominator does not make the failures imaginary, only unsuitable for estimating frequency by themselves.

Estimator has removed the every-bracket assumption and marked frequency and active time pending. Observation agreed, not completed yet.