@thegibson We spent probably 20 minutes writing a query to carefully select a subset of production servers upon which to execute a handful of *very* resource intensive commands, and then to disable a bunch of critical services.

We stored that data in a variable.

We referenced a different variable when we ran that command in our production datacenter.

It ran for about 10 minutes before I said "Gee, this is taking a really long time" and then my co-worker went in to a panic and reverted our changes.

Went from multiple million active users to less than 500k in about 5 minutes, back up to multiple millions in less than 3.

Happened so fast that it didn't even trigger an alert.