I accidentally nuked a Databricks Lakeflow Pipelines table that should never be deleted.
There was a problem in the pipeline, so I went to the pipeline and just hit full refresh. Turns out, the underlying data in storage was removed. This resulted in the loss of the full table history. Fortunately, Delta time travel was there to rescue me.
To avoid full refreshes on a Lakeflow Pipelines managed table, you can set the table property pipelines.reset.allowed to false.
So review your pipelines and ask yourself: what happens if someone hits full refresh?
