Latches are monitored to help identify performance bottlenecks.
Edit me

Total latch wait time is recalculated every seconds as the number of milliseconds spent in wait time for that second. A high rate of increase in total wait time (by 1000 or more ms/s) raises the Latch - Total Wait Time Rate alarm.

When the alarm is raised there has been a high increase (ms/s) in the total wait time for latch requests. Open the SQL Server | SQL Activity Drilldown | Locks page. Refer to the Latches chart for details.