[v3 migration 4/4] Verify the migrated release path inside the next real RC window

[os-project-manager]

Sub-issue of #9465.

Blocked-by: #9498

Restart-when: .changeset/pre.json exists on origin/main with "mode": "pre" — i.e. the next RC window has been entered.

On hold by design, not by neglect. Everything #9497 and #9498 measure happens in a throwaway clone or on a PR; the behaviour this epic exists to protect — a real cut, over a real accumulated changeset set, under the enforced cut-rc.yml allowlist — only exists inside a genuine pre-mode window. A migration that is green on a PR and wrong at cut time is exactly the failure mode we are buying insurance against, so the insurance is not claimed until it is testable.

What to check when it fires

• The first pnpm run version of the window: does it produce the file set [v3 migration 1/4] Rehearse @changesets/cli v3 in a throwaway clone and answer U1–U6 with measurements — no repo changes #9497 measured, and does cut-rc.yml's allowlist accept it without an override?
• Where consumed changesets actually go (.changeset/pre/ vs the root), and whether the count matches what the migrated gates expect.
• check-changeset-no-major.mjs against a live pre-mode repo: does its exemption behave as the migration assumed, and does a typo-sized PR in the window see a sane verdict rather than the 171-file absurdity the old shape produced?
• objectui-changeset-digest.mjs against the live pre.json.
• At window close: the post-exit state, which is where finding: check-changeset-no-major --self-test goes red repo-wide after a release exits pre mode — its control requires major-declaring changesets AND a pre.json that no longer exists #8654's self-test failure lived.

Non-goals

This card observes a release window; it does not create one. No pre-mode entry or exit, no publish, no tags, no release-workflow dispatch. If something is wrong, the deliverable is a precise report plus a fix card — not a hot fix applied mid-cut.

Acceptance

• A verification report on Migrate the release toolchain to @changesets/cli v3 — one atomic PR carrying the bump, the pre-mode restructure, and the gates that model v2's semantics #9465 covering each bullet above with real output from the window.
• Any divergence from [v3 migration 1/4] Rehearse @changesets/cli v3 in a throwaway clone and answer U1–U6 with measurements — no repo changes #9497's rehearsal filed as its own card with the measurement, and linked here — a rehearsal that mispredicted reality is a finding worth keeping, not an embarrassment to bury.