It'd be cool if cleanup could be told to keep the N last processed data sets of a component instead of only the latest.

Maybe add an ingestion timestamp to the database entries and use that a measure for how old a given data set is?

In kde neon we have some repos continuously integrate from git. This results in a large number of versions but out of those versions the N latest are conceivably "relevant", not only the latest. Currently we can't really use cleanup because it'd drop everything that isn't the current version and break screenshot resolution for systems that are still on the metadata from (e.g.) an hour ago.