Right now we only use h5py to append some floats to the .h5 checkpoint file. We really shouldn't be touching the HDF5 file for this purpose anyway. Instead we should either append the XML file or perhaps just have an extra pickle file.
As of now I'm not convinced that the current .h5 checkpoint file approach is wrong. It might be ideally to move towards representing an entire simulation in a .h5 file.
Right now we only use h5py to append some floats to the .h5 checkpoint file. We really shouldn't be touching the HDF5 file for this purpose anyway. Instead we should either append the XML file or perhaps just have an extra pickle file.