Memory leak problem of hiveserver2 when compiling in parallel

I used JDBC to connect to HIVESERVER2 and got about 25,000 SQL statements as test data from the production environment to test parallel compilation. Then save the memory snapshot of hiveserver2. Then run the test again and save the memory snapshot. I found that the memory occupied became larger and the second time running was more slower.

This is the first time I have submitted an issue. If there is any incomplete description, please point it out. If the format of the issue is incorrect, please help me modify it, thank you