Description
Currently the seed parameter is used for reproducibility, this means that two runs with same configuration, including seed, yield the same results.
There is a very important case that isn't covered right now.
Say we run the simulation with over 1E6 primaries, we perform some analysis on the events and end up with only a small subset (say 10) of events that are interesting to us (as it frequently happens in rare-event searches). We are able to reproduce these set of events from the initial configuration but we need to run the full simulation again, which is costly. It would be useful to have a way to only simulate the ones we know will be interesting to us, that way we could serialize an interesting but expensive simulation with only the configuration and it would be very cheap to reproduce the results back.
I guess a way to achieve this would be to add an optional input parameter with the event ID of the events we want to keep and stop those which aren't in the list early in the simulation. This would fail in MT mode. Perhaps there is a better way.