Settle the Subprocess-Failure Rule for a Status Printed in Message Text #3365

Context

The CODESTYLE.md rule proposed in #3363 classifies a subprocess failure by which command failed and its exit status, never by its message text. A local strict review pass on its pull request raised two follow-ups that fall outside that one-bullet change.

Scope Question

Some tools exit with one status for failures that need different handling. gh api exits 1 for every HTTP error, so a 404 (the object is absent, expected) and a 401 or a network failure (unexpected) share an exit status, and no split into separate operations gives the 404 a command of its own. The only signal separating them is the (HTTP 404) status code gh prints on stderr, which is a status code inside message text rather than prose.

Decide whether the rule admits a structured status a tool prints, such as an HTTP status code, and if so, how the bullet says it without enumerating tools.

Existing Sites

• scripts/pr_review.py matches HTTP_STATUS = re.compile(r"\(HTTP (\d{3})\)") against ABSENT = {"404", "422"}, the scope question above.
• scripts/pr_review.py retries a gh api call when its stderr contains "unknown flag: --allow-escape-sequences", which is version-dependent wording, the exact case the rule names.

Each site moves onto the rule in a change of its own once the scope question is answered.