We had an incident early this evening that resulted in 500 error responses and timeouts on our API. We've identified that this was triggered by a change we made earlier in the day and have rolled back that change. We're still investigating the underlying cause and the extent of the impact.