Healthcare admins need live dashboards without stale numbers. WebSockets carry appointment and schedule changes; MySQL remains the source of truth so every client eventually agrees after reconnect.
Reports that aggregate months of history do not belong on the request thread. Queued jobs generate exports and heavy summaries so the interactive UI keeps sub-second reads for day-to-day operations.
As appointment history grows, “slow query” often means missing or wrong indexes. We review execution plans against the filters we actually run–date ranges, clinic IDs, patient cohorts–and add covering indexes where it pays off.
The pattern is boring on purpose: real-time for what must be fresh, queues for what can wait, and queries that stay honest as data volume climbs.
