[leif]

I said most secondary indexes are high entropy.

If you have a large enough data set that you have to do a leaf read per insert (or close to that often), it will kill your throughput.