Constructing eras in an efficient manner is hard, and I've seen too many mistakes that are not obvious when you look at the era-construction algorithms.
I highly recommend adding some unit tests to make sure the end result is valid. Some tests that could easily be implemented:
Check that the person count per ingredient / condition including descendants hasn't changed
Check that for each person-ingredient/condition combination, the first start date hasn't changed
Check that for each drug / condition the number of eras <= the number of exposure / condition records
For each exposure / condition record, check that its start and end dates are included in an era
Constructing eras in an efficient manner is hard, and I've seen too many mistakes that are not obvious when you look at the era-construction algorithms.
I highly recommend adding some unit tests to make sure the end result is valid. Some tests that could easily be implemented:
I would personally also do a spot-check, for example using CohortDiagnostic's Cohort Explorer, but that may be my personal preference.