The code calls _graphql_span.__exit__(None, None, None) unconditionally in the finally block, which prevents exceptions from being recorded on the span. The Span.__exit__ method checks if value is not None to set SPANSTATUS.INTERNAL_ERROR, but since None is always passed for the exception value, any exceptions raised during GraphQL execution will result in spans without proper error status. This should use sys.exc_info() to pass actual exception information.

The __exit__(None, None, None) call always passes None for exception info regardless of whether an exception occurred during the GraphQL operation. The Span's __exit__ method checks if value is not None and should_be_treated_as_error(ty, value) to set SPANSTATUS.INTERNAL_ERROR, but since value is always None, exceptions during GraphQL operations will not mark the span as errored. This causes incorrect telemetry where failed operations appear successful.