Apache Druid

I don't have all the details, but coordinator does (at least) two things: balance and load.  If historicals are unbalanced, it might decide to move some, up to maxSegmentsToMove per cycle.  For loading (from deep storage), it will attempt to load up to maxSegmentsInNodeLoadingQueue per cycle.

My understanding is that it also prioritizes new segments (segments without any replicas in historicals) which would be the ones related to handoff from realtime ingestion.

all segments awaiting handoff would get processed by the relevant LoadRule on each execution of RunRules

the only throttling that applies to fresh segments is `maxSegmentsInNodeLoadingQueue` (if servers have very deep load queues, we'll wait til the next RunRules before adding more to load)