What is the best way to unit test a ML Pipeline

Hi all,

I’m looking for advice about how to properly unit test and maintain unit tests for a ML Pipeline where a team of Engineers needs to be able to add / remove features to the pipeline based on certain expected metrics.
The big problem is that development has slowed due to engineers having to constantly change unit tests further in the pipeline as features are added.

To the best of your knowledge:
What is your experience with testing / maintaining ML Pipelines?
What problems have you encountered and how did you overcome them?
How should the testing be structured?

I understand that this isn’t very specific but any advice will be appreciated!
Thank you!