My first question is why this matters? Is this curiosity or is there a real
performance issue you're tracking down?
I don't quite understand when you say machine A forwards...to machineB.
Are you talking about replication here? Or SolrCloud? Details matter, a lot
DIH has nothing that I know
I have found that segment flush is controlled by
DocumentWriterFlushControl, and indexing is implemented by
DocumentWriterPerThread. DocumentWriterFlushControl has information about
number of doc and size of RAM buffer, but this seemed be shared by
all DocumentWriterPerThread. Is that RAM limit is
