fix(federation/query-planner): Avoid unnecessarily precomputing non-trivial followup edges

.changesets/fix_sachin_avoid_precomputing_non_trivial_followup_edges_for_subgraph_query_graphs.md

@@ -0,0 +1,5 @@
+### Avoid unnecessary precomputation during query planner construction ([PR #8373](https://github.com/apollographql/router/pull/8373))
+
+A regression introduced in v2.5.0 caused query planner construction to unnecessarily precompute metadata, which can lead to increased CPU and memory utilization during supergraph loading. Query planner construction now correctly avoids this unnecessary precomputation.
+
+By [@sachindshinde](https://github.com/sachindshinde) in https://github.com/apollographql/router/pull/8373

apollo-federation/src/query_graph/build_query_graph.rs

@@ -427,8 +427,12 @@ impl SchemaQueryGraphBuilder {
         if self.for_query_planning {
             self.add_additional_abstract_type_edges()?;
         }
-        // This method adds no nodes/edges, but just precomputes followup edge information.
-        self.base.precompute_non_trivial_followup_edges()?;
+        // This method adds no nodes/edges, but just precomputes followup edge information. We don't
+        // need this when building subgraph query graphs (since federated query graph building takes
+        // care of it), so we accordingly skip precomputation in that case.
+        if self.subgraph.is_none() {
+            self.base.precompute_non_trivial_followup_edges()?;
+        }
         Ok(self.base.build())
     }