The error message says exactly what it means. If you try running the search repeatedly and keep getting this issue, then you may have an error in the connection path. For example, if you are on a multi-site clustered system, then perhaps the VPN link to the other site is wobbly, or perhaps some firewall in between is messing with the connection. Or, perhaps it is exactly what is says it might be, and the indexer restarted during the time that the search was running.
If this is happening to a job consistently at night, but not when you rerun during the day, try moving the job forward or backward to avoid the time that the indexer in question goes wonky.
In any case, this is not generally going to be a problem with your SPL (your search), rather it is some kind of problem with the architecture of your cluster or the timing of the job, relative to maintenance windows. See your system admins to ask them what they think it might be.
