Which transformation can reshape datasets without losing unique data?

The transformation that can reshape datasets without losing unique data is Union. The Union operation combines data from two or more datasets into a single dataset, ensuring that all unique records from each source are included in the final result. This means that if there are overlapping records in the datasets being combined, the Union will retain only one instance of each unique record, thereby preserving any unique data that exists across the input datasets.

In contrast, joins combine datasets based on common keys, which can lead to the loss of data if there are no matching records in the join keys. Coalesce is typically used to combine values from multiple fields into a single field, but it doesn't inherently deal with maintaining unique records across different datasets. Lastly, append simply adds one dataset to the end of another, which can lead to duplication if there are overlapping records without the same unique representation.

The Union transformation is specifically designed to handle the combination of multiple datasets while maintaining all unique entries, making it the correct choice for reshaping datasets without losing unique data.