Checkpoint files are dumps of all the information needed to restart a computation at the time of the checkpoint.
Some reasons you might want to do this:
For long simulations that take a lot of computer time, it’s good to write out checkpoint files periodically in case the run dies at some point (e.g. if you are running on a supercomputer with a time limit and you run out of time).
A checkpoint file written at the end of the simulation can be used to run out farther in time if you decide you need to.
You might want to restart the simulation at some intermediate time and specify more frequent outputs, if you something interesting happened that wasn’t captured in your original output times.
See Checkpointing and restarting in the Clawpack documentation for details on how to specify checkpoint intervals and how to restart from a checkpoint file.