[family collaborator] always read the freshest workspace state, even if my client tried to use a stale cache key

Prove that CareerVector resumes correctly after the things that actually happen in the field: a tab killed mid-write replays its IndexedDB queue, a browser dying mid-tailor leaves the process_request claimable by the CF Worker fallback, Durable Objects waking from hibernation continue serving the same Y.Doc bytes, edge caches stale relative to a fresh write are invalidated by clock, SSE clients catch up via Last-Event-Id, a corrupted IndexedDB cleanly falls back to the server snapshot, and a laptop sleeping mid-tailor resumes without duplicate writes.

Proof set for Edge cache stale after writes — clients refetch fresh sub-doc