When cluster shut down completely and restarted, new operations may be lost on client

In Geode, client keeps track of events received based on EventID. If duplicate events received from server, they are thrown away.

The current EventID takes parts of membership ID information, and it seems not adequate enough if whole cluster is down. (The coordinator is down and member viewId will start from 0 again.) This can lead to same event ids are generated, and cause client miss CQ events, etc.