SPA refetches the same endpoints redundantly on load (probe-stations 4x, energy/estimate 2x per space)

Captured the full network trace while walking all 11 SPA routes against the live brain. Several endpoints are fetched more times than the page needs. GET /settings/probe-stations was requested four times in immediate succession. GET /energy/estimate was requested twice for space 4×8 and twice for 2×4 with identical query strings. Separately, two polling loops are heavier than they look: /settings/hub-tunables is polled repeatedly and returns about 18 KB each time, and /fleet/computed polls every 5 seconds returning about 42 KB, which is a sustained 8 KB per second from a single browser tab. Two browser sessions were observed polling concurrently during the brain deploy, so that doubles. Nothing here is broken and no endpoint is slow enough to be visible on its own, but on a Pi serving a live grow this is avoidable load, and the duplicate identical requests suggest either several components each owning their own fetch or a missing request cache. Suggested: dedupe identical in-flight requests, share one subscription per endpoint across components, and consider a lighter delta shape for the fleet/computed poll. Found during the 8.1.0 full test pass.