From the discussion above, with good points on the value of data chunks containing multiple times, I'm starting to think that this is something to give the user of CMOR control over.
Would it be possible to add in a function call that allows us to set (a) the number of time records per chunk and (b) the total chunk size?
e.g. cmor_set_chunking(var_id, time_chunk_size, max_num_chunk_points) or cmor_set_chunking(var_id, chunk_shape_tuple)
The user would then be able to determine the chunking for the variable being written and optimise appropriately. In the longer term it would be valuable to provide guidance to groups on the optimum strategy.
Originally posted by @matthew-mizielinski in #880