Watch the size of your message queue!

Additionally watch the slow query log!

In order to make sure that updates are processed very fast, make sure that long running select only go a slave, so that the slow query log on the master for critical queries is always clean.