Production postmortem: Deduplicating replication speed

A customer called us with an interesting issue. They have a decently large database (around 750GB or so) that they want to replicate to another node. The did all the usual things that you need to do and the process started running as expected. However...