I was expecting crazy app abuses, but really the only one on the list was 100k connections. Instead it was crazy DB abuses, just as advertised.

I find the best way to handle database backups is to institute a -nightly DB living on a separate machine that you hand to your troubleshooting/QA/customer support team. It gets restored from backup every night. You will get your backup and restore in order doing that. It really helps if you make it useful in your infrastructure , by handing it out to people that can make use of it.

Our support team absolutely adores having a copy from yesterday basically always available.