In light of the recent changes to how we handle backend domain DNS changes (see #131), we now reload nginx whenever we detect a DNS change. Since we rolled that feature out, it's been working, but while debugging #188, I discovered that we were actually reloading nginx more frequently than we needed to. Basically, we were performing duplicate DNS queries in some scenarios, and in other race conditions we could send two reload commands in quick succession. This wasn't necessarily causing issues, but it was causing unnecessary reload churn on the server, which isn't ideal.