I have been using OpenSearch as a core component of the data plane for my customers specifically and exclusively for its:

  Search functions; and
  Data ingestion and transformation pipelines,

as well as a vector database for its k-NN approximate and radial similarity search functionality (with text embeddings for vector indices provided by another managed service). The current trench of work is focusing on moving all of the above into OpenSearch serverless collections.

I do not have the APM/monitoring use case anywhere near in my vicinity, and alarms and monitoring get griggered by / send metrics into CloudWatch.