Generalize the in-memory storage and adding data to it
The in-memory storage of flow slices should be somehow general ‒ something like a hash set or similar of the column values. The current tranformation there and back many times is uncomfortable and specific for the guts backend.
Furthermore, make it possible to assemble these informations from multiple sources. The sources of statistics can probably come from a single source, but the „headers“ should be composable.