The question is then mainly: when requests to that API fail, does your application workflow require that your clients/users
a) be notified of that
b) need to take a follow up action
c) is an HTTP error response the only way they can be notified of that?
If so: then consider when the remote API generates a 500 internal server error to have your ...