Starting around 4:10pm EDT on Wednesday, a large number of search events created a high load on our message queueing service. Unfortunately, that sustained high load caused the queuing service to fail. Help Scout services (web app, mobile API, help desk API, and the Docs API) were down for roughly 27 minutes, and services returned to normal around 4:40pm EDT.
We’ve made immediate changes to our infrastructure to keep this particular problem from happening again. Long term, we’re going to separate and expand our messaging queues, which will help distribute high loads more efficiently and redundantly going forward.