There's a reason I'm not a fan of curl. That it returns an exit code of 0 when a web site returns a 5xx error code by default. Yes, I can provide a -f to cause it to fail, but failng should be the default.

For reference, both wget and FreeBSD's fetch both return an error code by default.

(I'm also not a fan of the Python HTTP libraries that do the same. If the request fails, it should raise an exception, that is exactly what Python exceptions are there for. You can catch them and deal with the response that way instead of silently failing.)