TestFlight 1.0.2 (40) exposed missing series prices and unsuccessful grade lookups. The backend correction treats equivalent gallery denominators consistently, accepts an omitted ex suffix only with an exact collector number, shortens selected-grade history queries and serializes the existing bounded market queue. The client hydrates visible series tiles through authenticated Cardmarket-only batches, using cached eBay evidence without starting CompSniper searches. Selected grades follow slow queued work with bounded stored-only polling for about four minutes; there is still exactly one initiating live request.

Implementation: backend #511 and frontend #748.

The production audit found 27 requests used out of 10,000; five stored-only requests consumed zero provider calls. Effective limits remained 200 per rolling 24 hours, 10 per minute and 8,000 across the conservative monthly window. A single additional Giratina PSA 10 probe, behind the deployed PG budget, returned 15 rows, seven eligible French PSA 10 observations. Its latest qualified sale was 2025-01-30: EUR 518.96 at the dated ECB rate, explicitly an old-sale fallback.

Local provider fixtures and recorded packet replay do not establish exhaustive market coverage. Native iOS execution and a new TestFlight release remain pending; the correction is not part of build 40. No style, native, dependency or quota configuration changes are included.

Local checks: 1,289 API tests (zero skips), 553 ingest tests, API JS/TS checks, 2,698 frontend tests after integrating current staging REV-203 (11 existing skips), lint/typecheck and six paired real HTTP client/API fixture tests passed. The concurrent daily-cap regression allowed only one additional reservation from 199, stopping at 200. The two new/updated Maestro flows were parsed but not executed on a native device.