odd, reverting both changes singly seems fine, but when benchmarking together, reverting does have a small perf hit (on our synthetic microbenchmarks)

odd, reverting both changes singly seems fine, but when benchmarking together, reverting does have a small perf hit (on our synthetic microbenchmarks)

going to retain just the changes to the errors array rather than to the incrementalDataRecords array. the fix in terms of the latter is really to remove the entire idea of the GraphQLWrappedResult and instead filter incrementalDataRecords based on the bubbled null positions after execution, but this is something we will probably attempt only after incremental data delivery lands so that the implementation at this point can be closer to the spec.