In December 2025, I worked on an issue that had been around for some time.

At first, it looked like a problem with handling large datasets. While investigating it, I found that the actual issue came from a data transformation that had originally been added for the UI.

The UI needed to remove duplicate records for analytics so that the numbers were not counted incorrectly. Later, the same logic was also being used when returning data through the API. This caused some API users to miss records when they queried certain combinations of data.

The tricky part was that the original logic wasn't really wrong. It was doing what the UI needed. The problem was that the same logic was being applied to a different use case.

I added an option to allow API consumers to get the underlying data without that transformation.

The main thing I learned from this was that logic created for the UI shouldn't always be applied directly to API data. Something that works for one use case can cause problems when the same data is used somewhere else.