document / handle failed requests

[bnjmnrsh]

Currently, if any one of the API fetch calls fail, for example with 504 503 failures then these, being valid JSON responses will be passed to the client to handle locally.

Better document how to handle these errors, and or
Consider:

• providing a top-level convince object to check for errors without going through each call.
• can we 're-try' the request within the worker in the event of a failure, and stay within max execution time limits?