Environment

• graphql-core: v3.3.0a11
• strawberry-graphql: 0.x (with
  enable_experimental_incremental_execution=True)
• Python 3.12+

Description

When a query uses @stream_connection on a list field and the items in that list spread fragments with @defer, the incremental delivery pipeline deadlocks after delivering the initial result. The subsequent_results async generator never yields again and the HTTP response hangs indefinitely until
the server closes the connection.

e.g.

query MyListQuery {
    connection {
      edges @stream_connection(label: "MyList_edges", initialCount: 20) {
        node {
          ...MyItem_item @defer
        }
      }
   }
}

What happens

1. The initial response is delivered correctly (with hasNext:
   true)
2. subsequent_results begins iterating
3. No further values are yielded — the generator blocks
   forever
4. The HTTP connection eventually times out server-side

What we expect should happen

After the initial response, subsequent_results should yield the deferred fragments for each streamed edge, then the deferred pageInfo fragment, then a final { hasNext: false } terminator.

The main branch appears to include several incremental delivery fixes (e.g. commit 074a7c9 — "Finish async iterables in the non incremental delivery case") that may address this. However, the main branch has introduced API changes that are not yet compatible with downstream consumers like strawberry-graphql, so upgrading is not straightforward.