Prob just a coincidence, but our Memorystore (hosted redis) instance went down with a "repairing" status around the same time.