Eddy statistics aren't properly written when the code terminates unexpectedly
Example: when running on a cluster and the scheduler times out the job and kills it, the eddy stats files are not written.
Alternatively, we could open a file for reading and then append to it rather than writing everything at the end of the simulation.