The logserver collects and historizes statistics like profiling and monitoring data. This process happens regularly behind the scenes.
On systems with very complex queries, that contain thousands of profile parts and subselects, this historization may take a long time and consume significant computing resources. While this process is running, FLUSH STATISTICS commands are blocked and must wait for the process to finish.
The internal historization processes have been optimized, which improves the overall duration and resource usage of this process. Explicit FLUSH STATISTICS commands therefore are less likely to be blocked, and if they are, it is for a much shorter duration.