Test how long the example simulation takes, benchmark it.
See if caching with lru_cache has any impact whatsoever.
Get a sense of what time forward simulations will require.
Figure out how to collect input parameters in a way that is collapsable to a matrix of samples.