@mattleff Very interesting bug, once review comments are addressed I will merge it as a temporary fix.
But this uncovered generic issue since you can also reproduce with extendSchema or any other function that modify schema.
So the correct solution would be to just forbid returning GraphQLObjectType instance in the upcoming 16.0.0 so names of types would be the only option.
Do you see any potential problems with these approach?

@mattleff Thinking more about it it's better to have this temporary workaround directly inside execute:
https://github.com/graphql/graphql-js/blob/master/src/execution/execute.js#L1006-L1009
That way it's smaller and fix this problem for any schema transformation instead of just fixing lexicographicSortSchema

@IvanGoncharov Ok, I'll see if I can implement it in execute().

@IvanGoncharov I've pushed that change. Are you good with the current tests or is there a better place or method for testing this?

@mattleff Midnight in my location 🛏️
Will look into it tomorrow morning 🌅

@mattleff Took more time than I expected and resulted in #2790 #2791 #2793
So now we prepared for dropping GraphQLObject as a result of resolveType in the upcoming 16.0.0
Can you please test it with Next.js?

@mattleff You can use our NPM branch for tests:
https://github.com/graphql/graphql-js#want-to-ride-the-bleeding-edge

@IvanGoncharov That works great. Thanks for all your work to square this away!