When an operation is sent to the server we register the operation to timeout with the netty event loop. We properly unregister this when the task is completed normally. However if an exception we do not unregister the future which causes the operation to "leak" until the timeout can occur. Normally this defaults to 60 seconds but if this increased or there are a lot of operations or larger operations in that 60 second time frame it can cause memory issues. We need to stop the future even when there is an exception.