We had some problematic queries running in the database. Moreover, our monitoring process for problematic queries failed detecting them. The monitoring process is fixed and the queries causing congestion were killed. Now the system is no longer affected.