Generally you either have so many searches running that your Splunk can't keep up, search bundles too large because you have huge lookup files, etc. Or you disk I/o subsystem is too slow for one reason or another.
How many active jobs do you have in the job monitor?
Are you using NFS?
Are you seeing any other errors in index=_internal log_level=error ?
