Could this be considered a bug: https://github.com/delta-io/delta-rs/issues/1528 ? I noticed that on second writes the exact same pyarrow dataset couldn't be appended to the delta table due to it not casting the dtypes as it did on the first write.