Hello, my name is Vasilis Neris. I'm a magento fullstack developer with 4+ years of experience.
Your problem might be a bunch of things. Usually when all nodes die in a cluster, its usually due to memory. Since the previous freelancer tried that with no luck, maybe the best way to track the issue is to check what happens when the reindex is happening.
Possible Solutions:
Multi node setup on single cluster
Change Elasticsearch Version
I need to check logs etc to provide more solutions.