Just started reading a book called "High Performance MySQL" and in one of the early pages, the following advice appears:

"It's an excellent idea to run a realistic load simulation on a test server and then literally pull the power plug. The firsthand experience of recovering from a crash is priceless. It saves nasty surprises later."

Same goes for testing network connectivity and failover. I can't tell you how many times I've heard things like "The automatic recovery _should_ have kicked in but..."

Having a recovery procedure and backup strategy is completely different from having actually restored a backup and recovered from a failure.