Percent-encoding redirect test

A client site's /zh/ pages redirect an RFC-3986-normalized (uppercase) percent-encoded URL back to the site's native lowercase form — fine for a browser, but it sends Python's requests/urllib3 into an infinite redirect loop, because those libraries re-uppercase the redirect target on every hop.

Open question: does Googlebot do the same thing? Two isolated test pages, each logging every raw request it receives:

• Scenario t1 — a normal page with a correct, matching canonical. Its uppercase sibling isn't linked anywhere; if Google ever requests it anyway, that's Google re-casing on its own initiative.
• Scenario t2 — an identical page, except its own <link rel="canonical"> deliberately points at the redirecting uppercase form. This tests whether Google's own canonical-verification walks it into the same trap, independent of any re-casing behavior.

Both scenarios' uppercase forms return a 301 back to the lowercase original, exactly mirroring the client site's behavior.