The API went down for around 20 minutes this afternoon, from 6:06PM ET to 6:30M ET. This was due to an operator error: we had inadvertently introduced a software regression that caused the same database issue that occurred earlier in the week to resurface.

We will be discussing policy changes to prevent such regressions from occuring in the future, and our staff continues to investigate the source of the resource contention in order to resolve this issue permanently.