WhatsApp claims to use a well-known algorithm so it shouldn't be too hard for a researcher to encode a message with their own implementation of the algorithm (while copying the seed/keys from WhatsApp) and confirm that they get the same bytes out of both.
Of course that wouldn't prove that there's no backdoor, just that the normal code paths work as described.
Of course that wouldn't prove that there's no backdoor, just that the normal code paths work as described.