If we keep on gathering events indefinitely we will run out of memory. It is absolutely crucial to have a mechanism that prevents this (otherwise the live reduction will just crash).
Basically this boils down to:
monitoring memory consumption (depends on #11)
dropping events (depends on #14)
figuring out a reliable and robust way to combine those two, i.e., we have to make sure that the process reacts fast enough
If we keep on gathering events indefinitely we will run out of memory. It is absolutely crucial to have a mechanism that prevents this (otherwise the live reduction will just crash).
Basically this boils down to: