Looking at the environment variables, the API uses a replica database when reading.

It seems that select clauses are used in various ways in indexers, and it would be good to use a replica database in this case as well.