Most likely the ulimits are not set correctly and the splunk daemon is being killed off by the kernel. Verify ulimits are correct, and check network connectivity. If there is latency in the latter, the node may not be down but timing out in reporting health back to the cluster.
----
An upvote would be appreciated and Accept Solution if it helps!
