PR #11790 Cursor 4th re-review: manual granularity desyncs data tiers (CI watch + report)

Cursor Medium finding posted 2026-06-08T22:17:25Z on PR #11790 (winston/bolt-1193-brush-zoom-auto-densify). Fix pushed as 719651b at 22:48Z and DM sent to Slava in top-level DM at 22:51Z (Slack ts 1780958660.983969). This task tracks the CI watch + completion. Next heartbeat: 1. Re-check PR #11790 CI on 719651b. Expect Lint / Type Check / Unit Tests / Build / test/test / Shadow / Cursor Bugbot all SUCCESS within ~7-10 min. 2. If any required check fails, fix and re-push to winston/bolt-1193-brush-zoom-auto-densify. ASSUME-IMMEDIATE-MERGE applies (Slava self-merged the b02fe26 fix within an hour previously). 3. If all green and no further Cursor findings within 30 min, clear active-ci-watch.env and PATCH this task to completed. 4. If Cursor posts yet another finding, escalate as before. Fix summary: the timeRange-change useEffect in SiteEnergyCharts.tsx (lines ~282-318 after edit) now realigns BOTH windowSize and coarseWindowSize via getAppropriateWindowSize. Mirror of the site-change effect pattern. Trade-off: changing date range after manual granularity pick resets the manual choice (acceptable — two-tier contract > stale pick). New regression test in SiteEnergyCharts.test.tsx (after the site-change-zoom-null test) simulates the precondition: manual DAY pick via TimeControls dropdown → custom-picker range change firing ONLY onDateRangeChange → both windowSize tracks must snap back to HOUR. 27/27 SiteEnergyCharts tests pass locally. Lint + tsc clean.