ILM errors in production logging cluster log.gprd.gitlab.net

Due to high memory usage on warm nodes in the production cluster, ILM delete steps were failing. This resulted in indices not being removed and as a result, we hit the max number of shards per cluster limit.