Using Statement.add() and bind to execute batches is significantly slower than using Connection.createBatch(). Using bindings should be faster than recreating individual textual SQL statements and sending them over the connection.

As far as I have debugged it with wireshark, it looks like Connection.createBatch will send all the (mostly duplicated) SQL statements in a single packet back-to-back. On the other hand, statement batches do not send all of the bind/execute etc command in one go, but each statement is run one by one. Compare this to how JDBC statement-level batching works where all the binds and execute commands are also sent back-to-back in one go. For this reason (but not only), batched JDBC statements are also quite a bit faster than using r2dbc.

(In akka/akka-persistence-r2dbc@2ea4121 we were considering actual going to connection-level batching but issuing statements without bind (and creating individual textual SQL statements) seems to big of a risk / anti-pattern to work around this issue in the driver.)