Contentful organization → space → environment (+ environment aliases)

1. Leave de-DE price and image empty. GET the entry via the CDA in de-DE → expect the fallback-chain values to resolve.
2. Set de-DE price only. GET again → price is the DE override, image still inherited.
3. Update the master/default-locale price. GET de-DE → the inherited price reflects the new master value; the DE override is untouched.
4. Write an empty string "" into de-DE image. GET → show that this blocks the fallback ("what you set is what you get") — the documented trap.

Four CDA JSON payloads, side by side, showing steps 1–3 resolving correctly and step 4 breaking.

1. As an editor in the web app, attempt to publish a Draft entry → expect it blocked.
2. The decisive step: with a Management-API (CMA) token, POST a publish on that same Draft entry. Capture the HTTP response.
3. As a space admin, move the entry between steps → note that admins bypass restrictions by default (confirm how the admin role is contained).
4. Ask: can an integration/automation token be scoped so it, too, respects the workflow gate?

The raw CMA publish response from step 2 (a 2xx means the gate is UI-only; a 4xx means it is enforced).

1. Open the page preview; switch market/locale (DE ↔ another market) and desktop ↔ mobile.
2. Show which fields are inherited-from-HQ vs. locally overridden — visibly, in the editor, not just in the payload.
3. Show the entry's workflow state and a draft-vs-approved-vs-published comparison in the same flow.

A screen recording of the full flow, plus the plan gating for Studio (unverified — confirm).

1. Reference the external asset and render it through the delivery API + image transforms.
2. Show expiry / replacement handling and a 'where is this asset used' usage view.
3. Attempt to publish an entry pointing at the expired asset → expect a warning/block.

A working external-asset reference and a demonstrated (or clearly-scoped-as-custom) usage/expiry mechanism.

1. Run a re-runnable import (idempotent on an external key) twice → show no duplication.
2. Sustain CMA writes at the 10 req/s/space ceiling for ~10 min → capture the 429 / Retry-After behavior under load.
3. Clone an environment and swap the master alias → confirm what does NOT copy (workflows) and that versioning/snapshots exist only in the master-alias target.

Import logs (idempotency), a 429/Retry-After trace, and the alias-swap result.

1. Publish a change → measure publish-to-edge propagation time.
2. Ask for the concrete mainland-China delivery story (edge presence / ICP-compliant partner / customer-proxied).
3. Show the sanctioned cache-invalidation webhook/headers for BMW's Akamai — noting webhooks retry only 3× in ~1 min and never retry timeouts.

A propagation number and a written China-delivery architecture; the invalidation mechanism.

1. Run space export --download-assets → confirm content model, entries, assets, locales, roles and editor interfaces are all present.
2. Confirm the exclusions explicitly: workflows, tasks, scheduled releases, version history, space memberships, author history, webhook credentials.
3. Ask about the migration path for the proprietary rich-text JSON to another platform.

An export bundle + an explicit written list of what is and isn't included.

1. Get, in writing: which roadmap, pricing and renewal terms survive the Salesforce close, and any standalone-availability commitment.
2. Confirm the SLA ("up to 99.99%") as a contractual number, EU data-residency architecture (storage vs processing), and rate-limit uplifts for migration + steady state.
3. Confirm the custom-roles tier (Enterprise on today's pricing page) and the AI-Actions / bandwidth metering that apply to BMW's volumes.

Written answers destined for the contract/RFP — not verbal reassurance.