Architecture: should cli own baseline retry/backoff for safe requests?

[hawkfry]

Question: yes or no?

Should cli provide transport-level retry behavior for safe requests when failures are transient, such as:

• network errors
• 429
• 502
• 503
• 504

Value: this improves reliability for all callers by default, while leaving higher-level retry policy and workflow orchestration to the systems invoking the CLI.

If yes, I can implement this beneath the command layer.

[n8agrin]

Not opposed to this. I think the default should be no retry, but having something like a --retry optional flag seems useful