Better handling of time
Currently, each source holds its own internal time vector (computed from t0, dt, and size). However, if one is to write the response to a source in an existing file, this time vector is bypass, and that of the file is used.
This is super confusing and the source of many bugs or issues. We should change this behavior.