> I'd say the way failover/HA is done at Heroku is straightforward.

> It took some time to figure out because it required breaking some orthodoxy

If it's hard to figure out, it does not sounds too straight forward.