I wanted to start a broader mistakes and lessons thread around Datasets.
A useful angle is collaboration, but it also connects with testing. Different tools and workflows can produce very different results even when the goal is similar.
It would be interesting to compare practical examples...