Test optimistic-history reconciliation before a reload clears the evidence

When optimistic UI state is reconciled with persisted events through a pooled service, normalize both representations to the same canonical session identity within the authenticated owner/problem scope. A public lease ID and a worker-local session ID can refer to one logical turn while failing a session-plus-request-key join. Do not repair this by deduplicating text or dropping the session namespace: a request key can legitimately recur after a restart. Durable mappings should include stopped leases so history remains correct when workers are offline.

Test the overlap state itself: complete a turn, leave and reopen the view without a page reload, and assert exactly one prompt/reply with its activity preserved. Repeat hydration. Separately test identical text with distinct requests, request-key reuse across restarted sessions, and pending status. A .last() locator can tolerate duplicate rows, while a reload can erase the optimistic copy and make the test pass for the wrong reason.

Evidence/example — 2026-09-11: an authorized synthetic two-problem browser smoke showed duplicate completed rows after reopening a workspace, while scoped stored reviewer evidence contained one model execution. A deterministic pooled browser regression reproduced two matching DOM elements before reload; a small reconciliation reproduction yielded two rows before reload and one after clearing local state. This supports the failure mechanism and test method; it is not a claim that every identity scheme or deployment is covered. No candidate content, credentials, operational status, or raw logs are included.