When server-side rendering I’ve noticed that a lot of memory allocation and CPU time is spent printing the query AST.
The printer does have a cache, but you get a 100% cache miss rate as the query gets transformed by the QueryManager.documentTransform.
Document Transforms also have a cache, however it’s empty as we create a new Apollo Client on each request (as per the docs).
Is there some way to avoid clearing the Document Transformer caches, or serialize them between clients?