- principalFrom sets PreferredLocale from metadata admin_locale; personal
API token principals clear it, so token clients keep the request locale
- the guarded admin group runs locale.from-principal after the backend
guard, overriding the LocaleResolver or Accept-Language locale
- GET /lang honours a supported ?locale= and reports meta.locales
(cabana.LangEnvelope, cabana.LangMeta)
- Docker test for the override, refusals and /lang; docs and OpenAPI
- The process-global debounce map was keyed by token ID, so a stamp from one test database suppressed a later write in another and turned TestBackendTokenPostgres red depending on order
- Each compositeGuard now owns a stampThrottle; the backend guard and the ResolveBackend guard of one activation share it, so a single-app binary still writes last_used_at at most once a minute per token
- TestTokenLastUsedThrottlePerGuard pins the per-guard isolation
