And some different topic. I and my colleague Josef Handl (who is implementing the Kubernetes CSI driver) might found a problem in Oneclient. We were debugging a problem (lost connection) of Oneclient and we found a curious behaviour. This is probably caused by a bug in `helpers/src/communication/connectionPool.cc `. When the connection fails to restore, new attempt isn't scheduled, but the slot of pool remains blocked and when it runs out of all of them, communication isn't possible.