h5py 3.15: Diagnose Dataset Writes and Chunked I/O
Choose one answer. All questions are included, with explanations after checking. No time limit. Training target: 20/24. Applied diagnostics for numeric h5py 3.15 datasets: persistent writes, broadcasting, chunk geometry, compression, resizing, selection and bounded-memory reads. Assumes an open writable file unless stated. Excludes parallel HDF5, SWMR, virtual datasets, string encodings and storage latency guarantees. Sources: https://docs.h5py.org/en/3.15.0/high/dataset.html