In the context of Spark DataFrames, immutability refers to the creation of new data frames without modifying the original one. This principle is crucial for distributed processing, fault tolerance, and lineage tracking. It ensures data consistency across clusters, allows for reliable job reruns in case of failures, and simplifies understanding data transformations. Options A, C, and D describe scenarios that contradict the immutability principle, such as in-place modifications, which are not characteristic of Spark DataFrames.