The storage interruption took place between approximately 09:05h CEST and 10:05h CEST. It was caused by a corrupt storage node, which alternated between working and offline.
We identified the faulty node at around 09:20h CEST (the first assumption that it was a network issue delayed the identification) and tried to repair it. Towards 09:50h CEST it became clear that the node was not repairable and was replaced with a new one. During the whole event, the HTTP services for many Apps alternated between online and offline. About 15% of the Apps were fully offline for about 30 minutes and the rest between 5 and 10 minutes. New Apps were not affected at all.
We apologize for the inconvenience and will work on measurements to faster identify faulty nodes.