How to do ES fault drills

We want to do elasticsearch failure exercises, but currently we can only simulate some simple failures, such as network interruption and process death. The remaining deep-level failures are not well simulated. I would like to ask how the official elasticsearch developers simulate failures and conduct failure tests. Do you have any good suggestions in this regard?

You already asked a similar question at How to simulate the failure of elasticsearch?

Let's keep the discussion in a single place.

Thanks!