You are correct in realizing that if a query produces an error on the slave it will halt replication. However, this is only true IF the master succeeded on that same query. The MySQL binlog contains, among other things, the status of the query in question, so that if a query works fine on the master but the slave encounters duplicate entries, replication will halt. Also, queries such as the ones you mention, creating tables that already exist, or inserting into fields that don't exist, should never make it to the slave. I did some testing in my development replication environment, and none of what you posted panned out and replication was completely fine. I'm not sure what your problem is, but it honestly sounds like misconfiguration. :) Feel free to contact me if you have further questions.

Test comment