Comment by whakim

I did something very similar in my current job during my first few weeks. We had an "engineering test" script that monkeypatched the definition of "all users" (so that it returned the emails of the engineering team rather than the several hundred thousand people) before calling "send to everyone". This is obviously bad because someone might come along and change the internals of the "send to everyone" method and that person happened to be fresh-faced me. I had been tasked with making "send to everyone" a little more judicious (e.g. excluding people who hadn't opened an email in 90 days or had reported emails as spam), so I replaced the call to "all users" with some other method that calculated the recipients. And then, because I didn't totally understand how the script worked (and it seemed to be listing out some internal emails as recipients), I ran a final "engineering test" in production to see how it would look with real email content. Whoops x a few hundred thousand. I did get some pretty great responses to my apology email though.