My company has spent weeks agonizing over price changes and different features for different subscription levels (including a free level). We run all kinds of numbers and projections but at the end of the day you don't always know for sure until you flip the switch. You can do all kinds of surveys and talk to your customers. But you just don't know exactly how may people are going downgrade, upgrade, sign up, or even care about your changes.

The author was certainly a victim of their own success, as their conversion rates are pretty good. They just need to find limitations for the free account that doesn't consumer more resources than they can support.