Fair point.

This came from debugging a few production cases where everything looked fine at the API level, but delivery still varied depending on downstream handling.

Probably should’ve made that less abstract and more concrete.