The service was busy and didn't respond in time, was not a real offline scenario. The delay was caused in a spike of requests.