At the moment, one can only save a single microstate in an H5MD file at a time. For further statistical analysis, it would be convenient to have the entire trajectory stored to the same file.
The dump-function in DumpH5MDParallel.hpp:36 will then probably need the time step at which the data is supposed to be stored as input.