perf(timeseries): pre-aggregating user activity heatmaps with hourly UTC bucket rollups đź“Š
Instead of running
GROUP BY DATE(created_at AT TIME ZONE user_tz) across millions of raw event rows on every profile load, storing hourly UTC rollup counters lets you shift and render localized contribution graphs in sub-millisecond queries. How do you model multi-timezone analytics at scale?commit cfb062e