You could test for
1. correct error message pop ups that the connection is broken
2. test for whether all the processes in progress (in this case insertion, updates, deletion) were completed till the connection is broken
3. test for where the process needs to be restarted and how does the replication work
Basically we keep this type of testing under Exception Handling. These are various exceptions, which may arise during a process. You have to cater all components, which may fail during a process e.g.
1. Machine restarts
2. Network goes off
3. Hard disk of the server gets crashed (Do not know how to simulate)
4. Hard disk of the server gets full
5. Hard disk of the client gets full
6. After executing command, client is terminated abnormally; try to login again
In the similar fashion you can test other scenarios.
A highly advanced bug is indistinguishable from a feature.