we're seeing a lot of reads during write-heavy migration and it looks to be coming from schema cache misses. there are other areas we could make this more efficient but this is one of the things we're looking at.