Kinda related: sending a SIGSTOP to your datastore is a nice way to discover one of the failure modes of your architecture. (Use SIGCONT to make the server continue.)