really stress test DB replication

[binary1230]

I don't know the full extent here but this happened at Magfest 8.5

the internet went down for an entire day of registrations. when it came back, the slave server was out of sync with master.

this has something to do with a maximum size of the transaction logs (WAL logs) that can be sync'd without any intervention.

we should ensure that no matter how much data has changed, and however long the slave server has been offline, that the database is in a valid state on the slave after it sync's.

---

unrelated: a way to detect when this invalid state is occuring would be useful as well.

[binary1230]

mental note: investigate streaming replication as a solution to this.